backtesting

description: the process of testing a trading strategy or model on historical data

86 results

pages: 1,082 words: 87,792

Python for Algorithmic Trading: From Idea to Cloud Deployment
by Yves Hilpisch
Published 8 Dec 2020

Index A absolute maximum drawdown, Case Study AdaBoost algorithm, Vectorized Backtesting addition (+) operator, Data Types adjusted return appraisal ratio, Algorithmic Trading algorithmic trading (generally)advantages of, Algorithmic Trading basics, Algorithmic Trading-Algorithmic Trading strategies, Trading Strategies-Conclusions alpha seeking strategies, Trading Strategies alpha, defined, Algorithmic Trading anonymous functions, Python Idioms API key, for data sets, Working with Open Data Sources-Working with Open Data Sources Apple, Inc.intraday stock prices, Getting into the Basics reading stock price data from different sources, Reading Financial Data From Different Sources-Reading from Excel and JSON retrieving historical unstructured data about, Retrieving Historical Unstructured Data-Retrieving Historical Unstructured Data app_key, for Eikon Data API, Eikon Data API AQR Capital Management, pandas and the DataFrame Class arithmetic operations, Data Types array programming, Making Use of Vectorization(see also vectorization) automated trading operations, Automating Trading Operations-Strategy Monitoringcapital management, Capital Management-Kelly Criterion for Stocks and Indices configuring Oanda account, Configuring Oanda Account hardware setup, Setting Up the Hardware infrastructure and deployment, Infrastructure and Deployment logging and monitoring, Logging and Monitoring-Logging and Monitoring ML-based trading strategy, ML-Based Trading Strategy-Persisting the Model Object online algorithm, Online Algorithm-Online Algorithm Python environment setup, Setting Up the Python Environment Python scripts for, Python Script-Strategy Monitoring real-time monitoring, Real-Time Monitoring running code, Running the Code uploading code, Uploading the Code visual step-by-step overview, Visual Step-by-Step Overview-Real-Time Monitoring B backtestingbased on simple moving averages, Strategies Based on Simple Moving Averages-Generalizing the Approach Python scripts for classification algorithm backtesting, Classification Algorithm Backtesting Class Python scripts for linear regression backtesting class, Linear Regression Backtesting Class vectorized (see vectorized backtesting) BacktestLongShort class, Long-Short Backtesting Class, Long-Short Backtesting Class bar charts, matplotlib bar plots (see Plotly; streaming bar plot) base class, for event-based backtesting, Backtesting Base Class-Backtesting Base Class, Backtesting Base Class Bash script, Building a Ubuntu and Python Docker Imagefor Droplet set-up, Script to Orchestrate the Droplet Set Up-Script to Orchestrate the Droplet Set Up for Python/Jupyter Lab installation, Installation Script for Python and Jupyter Lab-Installation Script for Python and Jupyter Lab Bitcoin, pandas and the DataFrame Class, Working with Open Data Sources Boolean operationsNumPy, Boolean Operations pandas, Boolean Operations C callback functions, Retrieving Streaming Data capital managementautomated trading operations and, Capital Management-Kelly Criterion for Stocks and Indices Kelly criterion for stocks and indices, Kelly Criterion for Stocks and Indices-Kelly Criterion for Stocks and Indices Kelly criterion in binomial setting, Kelly Criterion in Binomial Setting-Kelly Criterion in Binomial Setting Carter, Graydon, FX Trading with FXCM CFD (contracts for difference)algorithmic trading risks, Logging and Monitoring defined, CFD Trading with Oanda risks of losses, Long-Short Backtesting Class risks of trading on margin, FX Trading with FXCM trading with Oanda, CFD Trading with Oanda-Python Script(see also Oanda) classification problemsmachine learning for, A Simple Classification Problem-A Simple Classification Problem neural networks for, The Simple Classification Problem Revisited-The Simple Classification Problem Revisited Python scripts for vectorized backtesting, Classification Algorithm Backtesting Class .close_all() method, Placing Orders cloud instances, Using Cloud Instances-Script to Orchestrate the Droplet Set Upinstallation script for Python and Jupyter Lab, Installation Script for Python and Jupyter Lab-Installation Script for Python and Jupyter Lab Jupyter Notebook configuration file, Jupyter Notebook Configuration File RSA public/private keys, RSA Public and Private Keys script to orchestrate Droplet set-up, Script to Orchestrate the Droplet Set Up-Script to Orchestrate the Droplet Set Up Cocteau, Jean, Building Classes for Event-Based Backtesting comma separated value (CSV) files (see CSV files) condaas package manager, Conda as a Package Manager-Basic Operations with Conda as virtual environment manager, Conda as a Virtual Environment Manager-Conda as a Virtual Environment Manager basic operations, Basic Operations with Conda-Basic Operations with Conda installing Miniconda, Installing Miniconda-Installing Miniconda conda remove, Basic Operations with Conda configparser module, The Oanda API containers (see Docker containers) contracts for difference (see CFD) control structures, Control Structures CPython, Python for Finance, Python Infrastructure .create_market_buy_order() method, Placing Orders .create_order() method, Placing Market Orders-Placing Market Orders cross-sectional momentum strategies, Strategies Based on Momentum CSV filesinput-output operations, Input-Output Operations-Input-Output Operations reading from a CSV file with pandas, Reading from a CSV File with pandas reading from a CSV file with Python, Reading from a CSV File with Python-Reading from a CSV File with Python .cummax() method, Case Study currency pairs, Logging and Monitoring(see also EUR/USD exchange rate) algorithmic trading risks, Logging and Monitoring D data science stack, Python, NumPy, matplotlib, pandas data snooping, Data Snooping and Overfitting data storageSQLite3 for, Storing Data with SQLite3-Storing Data with SQLite3 storing data efficiently, Storing Financial Data Efficiently-Storing Data with SQLite3 storing DataFrame objects, Storing DataFrame Objects-Storing DataFrame Objects TsTables package for, Using TsTables-Using TsTables data structures, Data Structures-Data Structures DataFrame class, pandas and the DataFrame Class-pandas and the DataFrame Class, Reading from a CSV File with pandas, DataFrame Class-DataFrame Class DataFrame objectscreating, Vectorization with pandas storing, Storing DataFrame Objects-Storing DataFrame Objects dataism, Preface DatetimeIndex() constructor, Plotting with pandas decision tree classification algorithm, Vectorized Backtesting deep learningadding features to analysis, Adding Different Types of Features-Adding Different Types of Features classification problem, The Simple Classification Problem Revisited-The Simple Classification Problem Revisited deep neural networks for predicting market direction, Using Deep Neural Networks to Predict Market Direction-Adding Different Types of Features market movement prediction, Using Deep Learning for Market Movement Prediction-Adding Different Types of Features trading strategies and, Machine and Deep Learning deep neural networks, Using Deep Neural Networks to Predict Market Direction-Adding Different Types of Features delta hedging, Algorithmic Trading dense neural network (DNN), The Simple Classification Problem Revisited, Using Deep Neural Networks to Predict Market Direction dictionary (dict) objects, Reading from a CSV File with Python, Data Structures DigitalOceancloud instances, Using Cloud Instances-Script to Orchestrate the Droplet Set Up droplet setup, Setting Up the Hardware DNN (dense neural network), The Simple Classification Problem Revisited, Using Deep Neural Networks to Predict Market Direction Docker containers, Using Docker Containers-Building a Ubuntu and Python Docker Imagebuilding a Ubuntu and Python Docker image, Building a Ubuntu and Python Docker Image-Building a Ubuntu and Python Docker Image defined, Docker Images and Containers Docker images versus, Docker Images and Containers Docker imagesdefined, Docker Images and Containers Docker containers versus, Docker Images and Containers Dockerfile, Building a Ubuntu and Python Docker Image-Building a Ubuntu and Python Docker Image Domingos, Pedro, Automating Trading Operations Droplet, Using Cloud Instancescosts, Infrastructure and Deployment script to orchestrate set-up, Script to Orchestrate the Droplet Set Up-Script to Orchestrate the Droplet Set Up dynamic hedging, Algorithmic Trading E efficient market hypothesis, Predicting Market Movements with Machine Learning Eikon Data API, Eikon Data API-Retrieving Historical Unstructured Dataretrieving historical structured data, Retrieving Historical Structured Data-Retrieving Historical Structured Data retrieving historical unstructured data, Retrieving Historical Unstructured Data-Retrieving Historical Unstructured Data Euler discretization, Python Versus Pseudo-Code EUR/USD exchange ratebacktesting momentum strategy on minute bars, Backtesting a Momentum Strategy on Minute Bars-Backtesting a Momentum Strategy on Minute Bars evaluation of regression-based strategy, Generalizing the Approach factoring in leverage/margin, Factoring In Leverage and Margin-Factoring In Leverage and Margin gross performance versus deep learning-based strategy, Using Deep Neural Networks to Predict Market Direction-Using Deep Neural Networks to Predict Market Direction, Adding Different Types of Features-Adding Different Types of Features historical ask close prices, Retrieving Historical Data-Retrieving Historical Data historical candles data for, Retrieving Candles Data historical tick data for, Retrieving Tick Data implementing trading strategies in real time, Implementing Trading Strategies in Real Time-Implementing Trading Strategies in Real Time logistic regression-based strategies, Generalizing the Approach placing orders, Placing Orders-Placing Orders predicting, Predicting Index Levels-Predicting Index Levels predicting future returns, Predicting Future Returns-Predicting Future Returns predicting index levels, Predicting Index Levels-Predicting Index Levels retrieving streaming data for, Retrieving Streaming Data retrieving trading account information, Retrieving Account Information-Retrieving Account Information SMA calculation, Getting into the Basics-Generalizing the Approach vectorized backtesting of ML-based trading strategy, Vectorized Backtesting-Vectorized Backtesting vectorized backtesting of regression-based strategy, Vectorized Backtesting of Regression-Based Strategy event-based backtesting, Building Classes for Event-Based Backtesting-Long-Short Backtesting Classadvantages, Building Classes for Event-Based Backtesting base class, Backtesting Base Class-Backtesting Base Class, Backtesting Base Class building classes for, Building Classes for Event-Based Backtesting-Long-Short Backtesting Class long-only backtesting class, Long-Only Backtesting Class-Long-Only Backtesting Class, Long-Only Backtesting Class long-short backtesting class, Long-Short Backtesting Class-Long-Short Backtesting Class, Long-Short Backtesting Class Python scripts for, Backtesting Base Class-Long-Short Backtesting Class Excelexporting financial data to, Exporting to Excel and JSON reading financial data from, Reading from Excel and JSON F featuresadding different types, Adding Different Types of Features-Adding Different Types of Features lags and, Using Logistic Regression to Predict Market Direction financial data, working with, Working with Financial Data-Python Scriptsdata set for examples, The Data Set Eikon Data API, Eikon Data API-Retrieving Historical Unstructured Data exporting to Excel/JSON, Exporting to Excel and JSON open data sources, Working with Open Data Sources-Working with Open Data Sources reading data from different sources, Reading Financial Data From Different Sources-Reading from Excel and JSON reading data from Excel/JSON, Reading from Excel and JSON reading from a CSV file with pandas, Reading from a CSV File with pandas reading from a CSV file with Python, Reading from a CSV File with Python-Reading from a CSV File with Python storing data efficiently, Storing Financial Data Efficiently-Storing Data with SQLite3 .flatten() method, matplotlib foreign exchange trading (see FX trading; FXCM) future returns, predicting, Predicting Future Returns-Predicting Future Returns FX trading, FX Trading with FXCM-References and Further Resources(see also EUR/USD exchange rate) FXCMFX trading, FX Trading with FXCM-References and Further Resources getting started, Getting Started placing orders, Placing Orders-Placing Orders retrieving account information, Account Information retrieving candles data, Retrieving Candles Data-Retrieving Candles Data retrieving data, Retrieving Data-Retrieving Candles Data retrieving historical data, Retrieving Historical Data-Retrieving Historical Data retrieving streaming data, Retrieving Streaming Data retrieving tick data, Retrieving Tick Data-Retrieving Tick Data working with the API, Working with the API-Account Information fxcmpy wrapper packagecallback functions, Retrieving Streaming Data installing, Getting Started tick data retrieval, Retrieving Tick Data fxTrade, CFD Trading with Oanda G GDX (VanEck Vectors Gold Miners ETF)logistic regression-based strategies, Generalizing the Approach mean-reversion strategies, Getting into the Basics-Generalizing the Approach regression-based strategies, Generalizing the Approach generate_sample_data(), Storing Financial Data Efficiently .get_account_summary() method, Retrieving Account Information .get_candles() method, Retrieving Historical Data .get_data() method, Backtesting Base Class, Retrieving Tick Data .get_date_price() method, Backtesting Base Class .get_instruments() method, Looking Up Instruments Available for Trading .get_last_price() method, Retrieving Streaming Data .get_raw_data() method, Retrieving Tick Data get_timeseries() function, Retrieving Historical Structured Data .get_transactions() method, Retrieving Account Information GLD (SPDR Gold Shares)logistic regression-based strategies, Using Logistic Regression to Predict Market Direction-Using Logistic Regression to Predict Market Direction mean-reversion strategies, Getting into the Basics-Generalizing the Approach gold pricemean-reversion strategies, Getting into the Basics-Getting into the Basics momentum strategy and, Getting into the Basics-Getting into the Basics, Generalizing the Approach-Generalizing the Approach Goldman Sachs, Python and Algorithmic Trading, Algorithmic Trading .go_long() method, Long-Short Backtesting Class H half Kelly criterion, Optimal Leverage Harari, Yuval Noah, Preface HDF5 binary storage library, Using TsTables-Using TsTables HDFStore wrapper, Storing DataFrame Objects-Storing DataFrame Objects high frequency trading (HFQ), Algorithmic Trading histograms, matplotlib hit ratio, defined, Vectorized Backtesting I if-elif-else control structure, Python Idioms in-sample fitting, Generalizing the Approach index levels, predicting, Predicting Index Levels-Predicting Index Levels infrastructure (see Python infrastructure) installation script, Python/Jupyter Lab, Installation Script for Python and Jupyter Lab-Installation Script for Python and Jupyter Lab Intel Math Kernel Library, Basic Operations with Conda iterations, Control Structures J JSONexporting financial data to, Exporting to Excel and JSON reading financial data from, Reading from Excel and JSON Jupyter Labinstallation script for, Installation Script for Python and Jupyter Lab-Installation Script for Python and Jupyter Lab RSA public/private keys for, RSA Public and Private Keys tools included, Using Cloud Instances Jupyter Notebook, Jupyter Notebook Configuration File K Kelly criterionin binomial setting, Kelly Criterion in Binomial Setting-Kelly Criterion in Binomial Setting optimal leverage, Optimal Leverage-Optimal Leverage stocks and indices, Kelly Criterion for Stocks and Indices-Kelly Criterion for Stocks and Indices Keras, Using Deep Learning for Market Movement Prediction, Using Deep Neural Networks to Predict Market Direction, Adding Different Types of Features key-value stores, Data Structures keys, public/private, RSA Public and Private Keys L lags, The Basic Idea for Price Prediction, Using Logistic Regression to Predict Market Direction lambda functions, Python Idioms LaTeX, Python Versus Pseudo-Code leveraged trading, risks of, Factoring In Leverage and Margin, FX Trading with FXCM, Optimal Leverage linear regressiongeneralizing the approach, Generalizing the Approach market movement prediction, Using Linear Regression for Market Movement Prediction-Generalizing the Approach predicting future market direction, Predicting Future Market Direction predicting future returns, Predicting Future Returns-Predicting Future Returns predicting index levels, Predicting Index Levels-Predicting Index Levels price prediction based on time series data, The Basic Idea for Price Prediction-The Basic Idea for Price Prediction review of, A Quick Review of Linear Regression scikit-learn and, Linear Regression with scikit-learn vectorized backtesting of regression-based strategy, Vectorized Backtesting of Regression-Based Strategy, Linear Regression Backtesting Class list comprehension, Python Idioms list constructor, Data Structures list objects, Reading from a CSV File with Python, Data Structures, Regular ndarray Object logging, of automated trading operations, Logging and Monitoring-Logging and Monitoring logistic regressiongeneralizing the approach, Generalizing the Approach-Generalizing the Approach market direction prediction, Using Logistic Regression to Predict Market Direction-Using Logistic Regression to Predict Market Direction Python script for vectorized backtesting, Classification Algorithm Backtesting Class long-only backtesting class, Long-Only Backtesting Class-Long-Only Backtesting Class, Long-Only Backtesting Class long-short backtesting class, Long-Short Backtesting Class-Long-Short Backtesting Class, Long-Short Backtesting Class longest drawdown period, Risk Analysis M machine learningclassification problem, A Simple Classification Problem-A Simple Classification Problem linear regression with scikit-learn, Linear Regression with scikit-learn market movement prediction, Using Machine Learning for Market Movement Prediction-Generalizing the Approach ML-based trading strategy, ML-Based Trading Strategy-Persisting the Model Object Python scripts, Linear Regression Backtesting Class trading strategies and, Machine and Deep Learning using logistic regression to predict market direction, Using Logistic Regression to Predict Market Direction-Using Logistic Regression to Predict Market Direction macro hedge funds, algorithmic trading and, Algorithmic Trading __main__ method, Backtesting Base Class margin trading, FX Trading with FXCM market direction prediction, Predicting Future Market Direction market movement predictiondeep learning for, Using Deep Learning for Market Movement Prediction-Adding Different Types of Features deep neural networks for, Using Deep Neural Networks to Predict Market Direction-Adding Different Types of Features linear regression for, Using Linear Regression for Market Movement Prediction-Generalizing the Approach linear regression with scikit-learn, Linear Regression with scikit-learn logistic regression to predict market direction, Using Logistic Regression to Predict Market Direction-Using Logistic Regression to Predict Market Direction machine learning for, Using Machine Learning for Market Movement Prediction-Generalizing the Approach predicting future market direction, Predicting Future Market Direction predicting future returns, Predicting Future Returns-Predicting Future Returns predicting index levels, Predicting Index Levels-Predicting Index Levels price prediction based on time series data, The Basic Idea for Price Prediction-The Basic Idea for Price Prediction vectorized backtesting of regression-based strategy, Vectorized Backtesting of Regression-Based Strategy market orders, placing, Placing Market Orders-Placing Market Orders math module, Data Types mathematical functions, Data Types matplotlib, matplotlib-matplotlib, Plotting with pandas-Plotting with pandas maximum drawdown, Risk Analysis, Case Study McKinney, Wes, pandas and the DataFrame Class mean-reversion strategies, NumPy and Vectorization, Strategies Based on Mean Reversion-Generalizing the Approachbasics, Getting into the Basics-Generalizing the Approach generalizing the approach, Generalizing the Approach Python code with a class for vectorized backtesting, Momentum Backtesting Class Miniconda, Installing Miniconda-Installing Miniconda mkl (Intel Math Kernel Library), Basic Operations with Conda ML-based strategies, ML-Based Trading Strategy-Persisting the Model Objectoptimal leverage, Optimal Leverage-Optimal Leverage persisting the model object, Persisting the Model Object Python script for, Automated Trading Strategy risk analysis, Risk Analysis-Risk Analysis vectorized backtesting, Vectorized Backtesting-Vectorized Backtesting MLPClassifier, The Simple Classification Problem Revisited MLTrader class, Online Algorithm-Online Algorithm momentum strategies, Momentumbacktesting on minute bars, Backtesting a Momentum Strategy on Minute Bars-Backtesting a Momentum Strategy on Minute Bars basics, Getting into the Basics-Getting into the Basics generalizing the approach, Generalizing the Approach Python code with a class for vectorized backtesting, Momentum Backtesting Class Python script for custom streaming class, Python Script Python script for momentum online algorithm, Momentum Online Algorithm vectorized backtesting of, Strategies Based on Momentum-Generalizing the Approach MomentumTrader class, Implementing Trading Strategies in Real Time-Implementing Trading Strategies in Real Time MomVectorBacktester class, Generalizing the Approach monitoringautomated trading operations, Logging and Monitoring-Logging and Monitoring, Real-Time Monitoring Python scripts for strategy monitoring, Strategy Monitoring Monte Carlo simulationsample tick data server, Sample Tick Data Server time series data based on, Python Scripts motives, for trading, Algorithmic Trading MRVectorBacktester class, Generalizing the Approach multi-layer perceptron, The Simple Classification Problem Revisited Musashi, Miyamoto, Python Infrastructure N natural language processing (NLP), Retrieving Historical Unstructured Data ndarray class, Vectorization with NumPy-Vectorization with NumPy ndarray objects, NumPy and Vectorization, ndarray Methods and NumPy Functions-ndarray Methods and NumPy Functionscreating, ndarray Creation linear regression and, A Quick Review of Linear Regression regular, Regular ndarray Object nested structures, Data Structures NLP (natural language processing), Retrieving Historical Unstructured Data np.arange(), ndarray Creation numbers, data typing of, Data Types numerical operations, pandas, Numerical Operations NumPy, NumPy and Vectorization-NumPy and Vectorization, NumPy-Random NumbersBoolean operations, Boolean Operations ndarray creation, ndarray Creation ndarray methods, ndarray Methods and NumPy Functions-ndarray Methods and NumPy Functions random numbers, Random Numbers regular ndarray object, Regular ndarray Object universal functions, ndarray Methods and NumPy Functions vectorization, Vectorization with NumPy-Vectorization with NumPy vectorized operations, Vectorized Operations numpy.random sub-package, Random Numbers NYSE Arca Gold Miners Index, Getting into the Basics O Oandaaccount configuration, Configuring Oanda Account account setup, Setting Up an Account API access, The Oanda API-The Oanda API backtesting momentum strategy on minute bars, Backtesting a Momentum Strategy on Minute Bars-Backtesting a Momentum Strategy on Minute Bars CFD trading, CFD Trading with Oanda-Python Script factoring in leverage/margin with historical data, Factoring In Leverage and Margin-Factoring In Leverage and Margin implementing trading strategies in real time, Implementing Trading Strategies in Real Time-Implementing Trading Strategies in Real Time looking up instruments available for trading, Looking Up Instruments Available for Trading placing market orders, Placing Market Orders-Placing Market Orders Python script for custom streaming class, Python Script retrieving account information, Retrieving Account Information-Retrieving Account Information retrieving historical data, Retrieving Historical Data-Factoring In Leverage and Margin working with streaming data, Working with Streaming Data Oanda v20 RESTful API, The Oanda API, ML-Based Trading Strategy-Persisting the Model Object, Vectorized Backtesting offline algorithmdefined, Signal Generation in Real Time transformation to online algorithm, Online Algorithm OLS (ordinary least squares) regression, matplotlib online algorithmautomated trading operations, Online Algorithm-Online Algorithm defined, Signal Generation in Real Time Python script for momentum online algorithm, Momentum Online Algorithm signal generation in real time, Signal Generation in Real Time-Signal Generation in Real Time transformation of offline algorithm to, Online Algorithm .on_success() method, Implementing Trading Strategies in Real Time, Online Algorithm open data sources, Working with Open Data Sources-Working with Open Data Sources ordinary least squares (OLS) regression, matplotlib out-of-sample evaluation, Generalizing the Approach overfitting, Data Snooping and Overfitting P package manager, conda as, Conda as a Package Manager-Basic Operations with Conda pandas, pandas and the DataFrame Class-pandas and the DataFrame Class, pandas-Input-Output OperationsBoolean operations, Boolean Operations case study, Case Study-Case Study data selection, Data Selection-Data Selection DataFrame class, DataFrame Class-DataFrame Class exporting financial data to Excel/JSON, Exporting to Excel and JSON input-output operations, Input-Output Operations-Input-Output Operations numerical operations, Numerical Operations plotting, Plotting with pandas-Plotting with pandas reading financial data from Excel/JSON, Reading from Excel and JSON reading from a CSV file, Reading from a CSV File with pandas storing DataFrame objects, Storing DataFrame Objects-Storing DataFrame Objects vectorization, Vectorization with pandas-Vectorization with pandas password protection, for Jupyter lab, Jupyter Notebook Configuration File .place_buy_order() method, Backtesting Base Class .place_sell_order() method, Backtesting Base Class Plotlybasics, The Basics multiple real-time streams for, Three Real-Time Streams multiple sub-plots for streams, Three Sub-Plots for Three Streams streaming data as bars, Streaming Data as Bars visualization of streaming data, Visualizing Streaming Data with Plotly-Streaming Data as Bars plotting, with pandas, Plotting with pandas-Plotting with pandas .plot_data() method, Backtesting Base Class polyfit()/polyval() convenience functions, matplotlib price prediction, based on time series data, The Basic Idea for Price Prediction-The Basic Idea for Price Prediction .print_balance() method, Backtesting Base Class .print_net_wealth() method, Backtesting Base Class .print_transactions() method, Retrieving Account Information pseudo-code, Python versus, Python Versus Pseudo-Code publisher-subscriber (PUB-SUB) pattern, Working with Real-Time Data and Sockets Python (generally)advantages of, Python for Algorithmic Trading basics, Python and Algorithmic Trading-References and Further Resources control structures, Control Structures data structures, Data Structures-Data Structures data types, Data Types-Data Types deployment difficulties, Python Infrastructure idioms, Python Idioms-Python Idioms NumPy and vectorization, NumPy and Vectorization-NumPy and Vectorization obstacles to adoption in financial industry, Python for Finance origins, Python for Finance pandas and DataFrame class, pandas and the DataFrame Class-pandas and the DataFrame Class pseudo-code versus, Python Versus Pseudo-Code reading from a CSV file, Reading from a CSV File with Python-Reading from a CSV File with Python Python infrastructure, Python Infrastructure-References and Further Resourcesconda as package manager, Conda as a Package Manager-Basic Operations with Conda conda as virtual environment manager, Conda as a Virtual Environment Manager-Conda as a Virtual Environment Manager Docker containers, Using Docker Containers-Building a Ubuntu and Python Docker Image using cloud instances, Using Cloud Instances-Script to Orchestrate the Droplet Set Up Python scriptsautomated trading operations, Running the Code, Python Script-Strategy Monitoring backtesting base class, Backtesting Base Class custom streaming class that trades a momentum strategy, Python Script linear regression backtesting class, Linear Regression Backtesting Class long-only backtesting class, Long-Only Backtesting Class long-short backtesting class, Long-Short Backtesting Class real-time data handling, Python Scripts-Sample Data Server for Bar Plot sample time series data set, Python Scripts strategy monitoring, Strategy Monitoring uploading for automated trading operations, Uploading the Code vectorized backtesting, Python Scripts-Mean Reversion Backtesting Class Q Quandlpremium data sets, Working with Open Data Sources working with open data sources, Working with Open Data Sources-Working with Open Data Sources R random numbers, Random Numbers random walk hypothesis, Predicting Index Levels range (iterator object), Control Structures read_csv() function, Reading from a CSV File with pandas real-time data, Working with Real-Time Data and Sockets-Sample Data Server for Bar PlotPython script for handling, Python Scripts-Sample Data Server for Bar Plot signal generation in real time, Signal Generation in Real Time-Signal Generation in Real Time tick data client for, Connecting a Simple Tick Data Client tick data server for, Running a Simple Tick Data Server-Running a Simple Tick Data Server, Sample Tick Data Server visualizing streaming data with Plotly, Visualizing Streaming Data with Plotly-Streaming Data as Bars real-time monitoring, Real-Time Monitoring Refinitiv, Eikon Data API relative maximum drawdown, Case Study returns, predicting future, Predicting Future Returns-Predicting Future Returns risk analysis, for ML-based trading strategy, Risk Analysis-Risk Analysis RSA public/private keys, RSA Public and Private Keys .run_mean_reversion_strategy() method, Long-Only Backtesting Class, Long-Short Backtesting Class .run_simulation() method, Kelly Criterion in Binomial Setting S S&P 500, Algorithmic Trading-Algorithmic Tradinglogistic regression-based strategies and, Generalizing the Approach momentum strategies, Getting into the Basics passive long position in, Kelly Criterion for Stocks and Indices-Kelly Criterion for Stocks and Indices scatter objects, Three Real-Time Streams scientific stack, NumPy and Vectorization, Python, NumPy, matplotlib, pandas scikit-learn, Linear Regression with scikit-learn ScikitBacktester class, Generalizing the Approach-Generalizing the Approach SciPy package project, NumPy and Vectorization seaborn library, matplotlib-matplotlib simple moving averages (SMAs), pandas and the DataFrame Class, Simple Moving Averagestrading strategies based on, Strategies Based on Simple Moving Averages-Generalizing the Approach visualization with price ticks, Three Real-Time Streams .simulate_value() method, Running a Simple Tick Data Server Singer, Paul, CFD Trading with Oanda sockets, real-time data and, Working with Real-Time Data and Sockets-Sample Data Server for Bar Plot sorting list objects, Data Structures SQLite3, Storing Data with SQLite3-Storing Data with SQLite3 SSL certificate, RSA Public and Private Keys storage (see data storage) streaming bar plots, Streaming Data as Bars, Sample Data Server for Bar Plot streaming dataOanda and, Working with Streaming Data visualization with Plotly, Visualizing Streaming Data with Plotly-Streaming Data as Bars string objects (str), Data Types-Data Types Swiss Franc event, CFD Trading with Oanda systematic macro hedge funds, Algorithmic Trading T TensorFlow, Using Deep Learning for Market Movement Prediction, Using Deep Neural Networks to Predict Market Direction Thomas, Rob, Working with Financial Data Thorp, Edward, Capital Management tick data client, Connecting a Simple Tick Data Client tick data server, Running a Simple Tick Data Server-Running a Simple Tick Data Server, Sample Tick Data Server time series data setspandas and vectorization, Vectorization with pandas price prediction based on, The Basic Idea for Price Prediction-The Basic Idea for Price Prediction Python script for generating sample set, Python Scripts SQLite3 for storage of, Storing Data with SQLite3-Storing Data with SQLite3 TsTables for storing, Using TsTables-Using TsTables time series momentum strategies, Strategies Based on Momentum(see also momentum strategies) .to_hdf() method, Storing DataFrame Objects tpqoa wrapper package, The Oanda API, Working with Streaming Data trading platforms, factors influencing choice of, CFD Trading with Oanda trading strategies, Trading Strategies-Conclusions(see also specific strategies) implementing in real time with Oanda, Implementing Trading Strategies in Real Time-Implementing Trading Strategies in Real Time machine learning/deep learning, Machine and Deep Learning mean-reversion, NumPy and Vectorization momentum, Momentum simple moving averages, Simple Moving Averages trading, motives for, Algorithmic Trading transaction costs, Long-Only Backtesting Class, Vectorized Backtesting TsTables package, Using TsTables-Using TsTables tuple objects, Data Structures U Ubuntu, Building a Ubuntu and Python Docker Image-Building a Ubuntu and Python Docker Image universal functions, NumPy, ndarray Methods and NumPy Functions V v20 wrapper package, The Oanda API, ML-Based Trading Strategy-Persisting the Model Object, Vectorized Backtesting value-at-risk (VAR), Risk Analysis-Risk Analysis vectorization, NumPy and Vectorization, Strategies Based on Mean Reversion-Generalizing the Approach vectorized backtestingdata snooping and overfitting, Data Snooping and Overfitting-Conclusions ML-based trading strategy, Vectorized Backtesting-Vectorized Backtesting momentum-based trading strategies, Strategies Based on Momentum-Generalizing the Approach potential shortcomings, Building Classes for Event-Based Backtesting Python code with a class for vectorized backtesting of mean-reversion trading strategies, Momentum Backtesting Class Python scripts for, Python Scripts-Mean Reversion Backtesting Class, Linear Regression Backtesting Class regression-based strategy, Vectorized Backtesting of Regression-Based Strategy trading strategies based on simple moving averages, Strategies Based on Simple Moving Averages-Generalizing the Approach vectorization with NumPy, Vectorization with NumPy-Vectorization with NumPy vectorization with pandas, Vectorization with pandas-Vectorization with pandas vectorized operations, Vectorized Operations virtual environment management, Conda as a Virtual Environment Manager-Conda as a Virtual Environment Manager W while loops, Control Structures Z ZeroMQ, Working with Real-Time Data and Sockets About the Author Dr.

…

Quantitative Trading: How to Build Your Own Algorithmic Trading Business
by Ernie Chan
Published 17 Nov 2008

In this chapter, I will describe the common platforms that can be used for backtesting, various sources of historical data useful for backtesting, a minimal set of standard performance measures that a backtest should provide, common pitfalls to avoid, and simple refinements and improvements to strategies. A few fully developed backtesting examples will also be presented to illustrate the principles and techniques described. 31 P1: JYS c03 JWBK321-Chan September 24, 2008 13:52 Printer: Yet to come 32 QUANTITATIVE TRADING COMMON BACKTESTING PLATFORMS There are numerous commercial platforms that are designed for backtesting, some of them costing tens of thousands of dollars.

…

24 vii P1: JYS fm JWBK321-Chan September 24, 2008 13:43 Printer: Yet to come viii Does the Strategy Suﬀer from Data-Snooping Bias? Does the Strategy “Fly under the Radar" of Institutional Money Managers? CONTENTS 25 27 Summary 28 CHAPTER 3 Backtesting 31 Common Backtesting Platforms 32 Excel 32 MATLAB 32 TradeStation 35 High-End Backtesting Platforms 35 Finding and Using Historical Databases 36 Are the Data Split and Dividend Adjusted? 36 Are the Data Survivorship Bias Free? 40 Does Your Strategy Use High and Low Data? 42 Performance Measurement 43 Common Backtesting Pitfalls to Avoid 50 Look-Ahead Bias 51 Data-Snooping Bias 52 Transaction Costs 60 Strategy Reﬁnement 65 Summary 66 CHAPTER 4 Setting Up Your Business 69 Business Structure: Retail or Proprietary?

…

Nevertheless, it is a simple illustration of MATLAB’s text processing functions. P1: JYS c03 JWBK321-Chan September 24, 2008 13:52 Backtesting Printer: Yet to come 35 TradeStation TradeStation (www.tradestation.com) is familiar to many retail traders as a brokerage that provides all-in-one backtesting and trade execution platforms linked to the brokerage’s servers. The main advantages of this setup are: r Most of the historical data necessary for backtesting is readily available, whereas you have to download the data from somewhere else if you use Excel or MATLAB. r Once you have backtested the program, you can immediately generate orders using the same program and transmit them to the brokerage.

pages: 571 words: 105,054

Advances in Financial Machine Learning
by Marcos Lopez de Prado
Published 2 Feb 2018

CHAPTER 11 The Dangers of Backtesting 11.1 Motivation Backtesting is one of the most essential, and yet least understood, techniques in the quant arsenal. A common misunderstanding is to think of backtesting as a research tool. Researching and backtesting is like drinking and driving. Do not research under the influence of a backtest. Most backtests published in journals are flawed, as the result of selection bias on multiple tests (Bailey, Borwein, López de Prado, and Zhu [2014]; Harvey et al. [2016]). A full book could be written listing all the different errors people make while backtesting. I may be the academic author with the largest number of journal articles on backtesting1 and investment performance metrics, and still I do not feel I would have the stamina to compile all the different errors I have seen over the past 20 years.

…

If you do that, the chances of finding a false discovery will drop substantially, but still they will not be zero. 11.5 A Few General Recommendations Backtest overfitting can be defined as selection bias on multiple backtests. Backtest overfitting takes place when a strategy is developed to perform well on a backtest, by monetizing random historical patterns. Because those random patterns are unlikely to occur again in the future, the strategy so developed will fail. Every backtested strategy is overfit to some extent as a result of “selection bias”: The only backtests that most people share are those that portray supposedly winning investment strategies. How to address backtest overfitting is arguably the most fundamental question in quantitative finance.

…

Stacked Feature Importance 8.6 Experiments with Synthetic Data Exercises References Note Chapter 9 Hyper-Parameter Tuning with Cross-Validation 9.1 Motivation 9.2 Grid Search Cross-Validation 9.3 Randomized Search Cross-Validation 9.4 Scoring and Hyper-parameter Tuning Exercises References Bibliography Notes PART 3 BACKTESTING Chapter 10 Bet Sizing 10.1 Motivation 10.2 Strategy-Independent Bet Sizing Approaches 10.3 Bet Sizing from Predicted Probabilities 10.4 Averaging Active Bets 10.5 Size Discretization 10.6 Dynamic Bet Sizes and Limit Prices Exercises References Bibliography Notes Chapter 11 The Dangers of Backtesting 11.1 Motivation 11.2 Mission Impossible: The Flawless Backtest 11.3 Even If Your Backtest Is Flawless, It Is Probably Wrong 11.4 Backtesting Is Not a Research Tool 11.5 A Few General Recommendations 11.6 Strategy Selection Exercises References Bibliography Note Chapter 12 Backtesting through Cross-Validation 12.1 Motivation 12.2 The Walk-Forward Method 12.3 The Cross-Validation Method 12.4 The Combinatorial Purged Cross-Validation Method 12.5 How Combinatorial Purged Cross-Validation Addresses Backtest Overfitting Exercises References Chapter 13 Backtesting on Synthetic Data 13.1 Motivation 13.2 Trading Rules 13.3 The Problem 13.4 Our Framework 13.5 Numerical Determination of Optimal Trading Rules 13.6 Experimental Results 13.7 Conclusion Exercises References Notes Chapter 14 Backtest Statistics 14.1 Motivation 14.2 Types of Backtest Statistics 14.3 General Characteristics 14.4 Performance 14.5 Runs 14.6 Implementation Shortfall 14.7 Efficiency 14.8 Classification Scores 14.9 Attribution Exercises References Bibliography Notes Chapter 15 Understanding Strategy Risk 15.1 Motivation 15.2 Symmetric Payouts 15.3 Asymmetric Payouts 15.4 The Probability of Strategy Failure Exercises References Chapter 16 Machine Learning Asset Allocation 16.1 Motivation 16.2 The Problem with Convex Portfolio Optimization 16.3 Markowitz's Curse 16.4 From Geometric to Hierarchical Relationships 16.5 A Numerical Example 16.6 Out-of-Sample Monte Carlo Simulations 16.7 Further Research 16.8 Conclusion APPENDICES 16.A.1 Correlation-based Metric 16.A.2 Inverse Variance Allocation 16.A.3 Reproducing the Numerical Example 16.A.4 Reproducing the Monte Carlo Experiment Exercises References Notes PART 4 USEFUL FINANCIAL FEATURES Chapter 17 Structural Breaks 17.1 Motivation 17.2 Types of Structural Break Tests 17.3 CUSUM Tests 17.4 Explosiveness Tests Exercises References Chapter 18 Entropy Features 18.1 Motivation 18.2 Shannon's Entropy 18.3 The Plug-in (or Maximum Likelihood) Estimator 18.4 Lempel-Ziv Estimators 18.5 Encoding Schemes 18.6 Entropy of a Gaussian Process 18.7 Entropy and the Generalized Mean 18.8 A Few Financial Applications of Entropy Exercises References Bibliography Note Chapter 19 Microstructural Features 19.1 Motivation 19.2 Review of the Literature 19.3 First Generation: Price Sequences 19.4 Second Generation: Strategic Trade Models 19.5 Third Generation: Sequential Trade Models 19.6 Additional Features from Microstructural Datasets 19.7 What Is Microstructural Information?

Learn Algorithmic Trading
by Sebastien Donadio
Published 7 Nov 2019

While the code of a for-loop backtester is easy to write, we should still use this type of backtester to eliminate low-performance strategies. If a strategy does not perform well with for-loop backtesters, this means that it will perform even worse on more realistic backtesters. Since it is important to have a backtester that's as realistic as possible, we will learn how an event-driven backtester works in the following section. Event-driven backtest systems An event-driven backtester uses almost all the components of the trading system. Most of the time, this type of backtester encompass all the trading system components (such as the order manager system, the position manager, and the risk manager).

…

Choice of IDE – Pycharm or Notebook Our first algorithmic trading (buy when the price is low, and sell when the price is high) Setting up your workspace PyCharm 101 Getting the data Preparing the data – signal Signal visualization Backtesting Summary Section 2: Trading Signal Generation and Strategies Deciphering the Markets with Technical Analysis Designing a trading strategy based on trend- and momentum-based indicators Support and resistance indicators Creating trading signals based on fundamental technical analysis Simple moving average Implementation of the simple moving average Exponential moving average Implementation of the exponential moving average Absolute price oscillator Implementation of the absolute price oscillator Moving average convergence divergence Implementation of the moving average convergence divergence Bollinger bands Implementation of Bollinger bands Relative strength indicator Implementation of the relative strength indicator Standard deviation Implementing standard derivatives Momentum Implementation of momentum Implementing advanced concepts, such as seasonality, in trading instruments Summary Predicting the Markets with Basic Machine Learning Understanding the terminology and notations Exploring our financial dataset Creating predictive models using linear regression methods Ordinary Least Squares Regularization and shrinkage – LASSO and Ridge regression Decision tree regression Creating predictive models using linear classification methods K-nearest neighbors Support vector machine Logistic regression Summary Section 3: Algorithmic Trading Strategies Classical Trading Strategies Driven by Human Intuition Creating a trading strategy based on momentum and trend following Examples of momentum strategies Python implementation Dual moving average Naive trading strategy Turtle strategy Creating a trading strategy that works for markets with reversion behavior Examples of reversion strategies Creating trading strategies that operate on linearly correlated groups of trading instruments Summary Sophisticated Algorithmic Strategies Creating a trading strategy that adjusts for trading instrument volatility Adjusting for trading instrument volatility in technical indicators Adjusting for trading instrument volatility in trading strategies Volatility adjusted mean reversion trading strategies Mean reversion strategy using the absolute price oscillator trading signal Mean reversion strategy that dynamically adjusts for changing volatility Trend-following strategy using absolute price oscillator trading signal Trend-following strategy that dynamically adjusts for changing volatility Creating a trading strategy for economic events Economic releases Economic release format Electronic economic release services Economic releases in trading Understanding and implementing basic statistical arbitrage trading strategies Basics of StatArb Lead-lag in StatArb Adjusting portfolio composition and relationships Infrastructure expenses in StatArb StatArb trading strategy in Python StatArb data set Defining StatArb signal parameters Defining StatArb trading parameters Quantifying and computing StatArb trading signals StatArb execution logic StatArb signal and strategy performance analysis Summary Managing the Risk of Algorithmic Strategies Differentiating between the types of risk and risk factors Risk of trading losses Regulation violation risks Spoofing Quote stuffing Banging the close Sources of risk Software implementation risk DevOps risk Market risk Quantifying the risk The severity of risk violations Differentiating the measures of risk Stop-loss Max drawdown Position limits Position holding time Variance of PnLs Sharpe ratio Maximum executions per period Maximum trade size Volume limits Making a risk management algorithm Realistically adjusting risk Summary  Section 4: Building a Trading System Building a Trading System in Python Understanding the trading system Gateways Order book management Strategy Order management system  Critical components Non-critical components Command and control Services Building a trading system in Python LiquidityProvider class Strategy class OrderManager class MarketSimulator class TestTradingSimulation class Designing a limit order book Summary Connecting to Trading Exchanges Making a trading system trade with exchanges Reviewing the Communication API Network basics Trading protocols FIX communication protocols Price updates Orders Receiving price updates Initiator code example Price updates Sending orders and receiving a market response Acceptor code example Market Data request handling Order Other trading APIs Summary Creating a Backtester in Python Learning how to build a backtester  In-sample versus out-of-sample data Paper trading (forward testing) Naive data storage HDF5 file Databases Relational databases Non-relational databases Learning how to choose the correct assumptions For-loop backtest systems Advantages Disadvantages Event-driven backtest systems Advantages Disadvantages Evaluating what the value of time is Backtesting the dual-moving average trading strategy For-loop backtester Event-based backtester Summary Section 5: Challenges in Algorithmic Trading Adapting to Market Participants and Conditions Strategy performance in backtester versus live markets Impact of backtester dislocations Signal validation Strategy validation Risk estimates Risk management system Choice of strategies for deployment Expected performance Causes of simulation dislocations Slippage Fees Operational issues Market data issues Latency variance Place-in-line estimates Market impact Tweaking backtesting and strategies in response to live trading Historical market data accuracy Measuring and modeling latencies Improving backtesting sophistication Adjusting expected performance for backtester bias Analytics on live trading strategies Continued profitability in algorithmic trading Profit decay in algorithmic trading strategies Signal decay due to lack of optimization Signal decay due to absence of leading participants Signal discovery by other participants Profit decay due to exit of losing participants Profit decay due to discovery by other participants Profit decay due to changes in underlying assumptions/relationships Seasonal profit decay Adapting to market conditions and changing participants Building a trading signals dictionary/database Optimizing trading signals Optimizing prediction models Optimizing trading strategy parameters Researching new trading signals Expanding to new trading strategies Portfolio optimization Uniform risk allocation PnL-based risk allocation PnL-sharpe-based risk allocation Markowitz allocation Regime Predictive allocation Incorporating technological advances Summary Final words Other Books You May Enjoy Leave a review - let other readers know what you think Preface In modern times, it is increasingly difficult to gain a significant competitive edge just by being faster than others, which means relying on sophisticated trading signals, predictive models, and strategies.

…

Once a model is implemented, it is necessary to test whether the trading robot behaves as expected in the trading infrastructure. In this chapter, we will learn how backtesting works, and then we will talk about the assumptions you will need to consider when creating a backtester. Finally, we will provide a backtester example by using a momentum trading strategy. In this chapter, we will cover the following topics: Learning how to build a backtester Learning how to choose the correct assumptions Evaluating what the value of time is Backtesting the dual-moving average trading strategy Learning how to build a backtester Backtesting is key in the creation of trading strategies. It assesses how profitable a trading strategy is by using historical data.

Systematic Trading: A Unique New Method for Designing Trading and Investing Systems
by Robert Carver
Published 13 Sep 2015

Now you’re using Sharpe ratios (SR) to produce your handcrafted weights it’s worth reiterating that this is a mild form of in-sample back-test cheating, since you only use the final SR averaged over all data history, which you wouldn’t have at the beginning of the back-test.68 Again this is a fair criticism, but the problem is not that serious. The weights are still not extreme, so the effect on back-tested SR you get is modest compared to in-sample single period optimisation. However you should still be cautious of assuming that you’d be able to achieve the back-test SR in live trading. Table 14 shows you roughly how much you should degrade back-tested returns to get realistic achievable Sharpe ratios given a particular fitting technique for a system like the one I describe in chapter fifteen. 68.

…

Recommended percentage volatility targets I run a highly diversified futures trading system with around 45 instruments, eight trading rules drawn from four different styles, and 30 trading rule variations. In a 35 year back-test, conservatively fitted with out of sample bootstrapping, it has a Sharpe ratio (SR) of around 1.0 after costs, but the highest volatility target I’d advocate using 145 Systematic Trading for it is 37%, rather than the 100% suggested by the Kelly criterion and the back-tested SR.106 Why such a conservative number – am I a wimp? There are several reasons for my caution. Firstly, it’s unlikely a back-tested SR will be achieved in the future. On average realised performance is never as good as in back-tests. This is because it’s very easy to over-fit if you ignore the advice in chapters three and four.

…

Estimating the number of round trips You can now calculate the expected costs of trading each instrument, given the units of turnover of the trading system and the standardised costs. But how do you find the turnover? There are three ways to estimate it. Sophisticated backtest If you have access to decent back-testing software, or can write your own, then you can include a function that estimates turnover directly. Simple back-test More rudimentary back-testing tools can give you an estimate of the average number of instrument blocks traded each year and the average absolute number of blocks you held. You can then calculate turnover as: Average number of blocks traded per year ÷ (2 × average absolute number of blocks held) Rule of thumb The third alternative is to use rules of thumb.

pages: 321

Finding Alphas: A Quantitative Approach to Building Trading Strategies
by Igor Tulchinsky
Published 30 Sep 2019

In a time-series holdout, researchers do not conduct the backtest on a given section of time. Similarly, in an asset holdout researchers do not conduct the backtest on a certain set of assets. After the holdout is incorporated into the backtest, a separate backtest is conducted solely on the holdout to validate whether the performance is consistent. There are two common approaches when using a time-series holdout: 1. Omit a continuous stretch of time. Typically, this is toward the end of the time series. 2. Hold out several interleaved stretches of time within the entire backtesting window. For example, hold out periods every alternate week within that window.

…

An automated system lacks this advantage, so one may want to consider quantitative methods for detecting when the backtesting period is too long, such as splitting the backtest and checking consistency across periods. At the cost of a modest increase in computational complexity, it may be possible to update certain parameters dynamically within the alphas rather than fitting them as part of the search, so that a longer backtesting period remains relevant. Alphas from Automated Search117 Another concern with a longer backtesting period, especially in a large-scale search, is the higher computational load. In an iterative search, an incremental backtest period is a useful trick to take advantage of a longer backtest period without using excessive resources.

…

The optimal ratio of in-sample to out-of-sample data in model building depends on the model’s complexity. LOOKING BACK Backtesting involves looking back in time to evaluate how a forecast or trading strategy would have performed historically. Although backtesting is invaluable (providing a window into both the markets and how the alpha would have performed), there are two important points to remember: •• History does not repeat itself exactly. So while an alpha idea may look great in a backtest, there’s no guarantee (only a level of confidence) it will continue to work in the future. This is because of the perverse power of computation and the ability of creative modelers 14 Finding Alphas to miss the forest for the trees.

Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals
by David Aronson
Published 1 Nov 2006

This is so because the performance of a rule can be profoundly affected by factors that have nothing to do with its predictive power. The Conjoint Effect of Position Bias and Market Trend on Back-Test Performance In reality, a rule’s back-tested performance is comprised of two independent components. One component is attributable to the rule’s predictive power, if it has any. This is the component of interest. The second, and unwanted, component of performance is the result of two factors that have nothing to do with the rule’s predictive power: (1) the rule’s long/short position bias, and (2) the market’s net trend during the back-test period. This undesirable component of performance can dramatically inﬂuence 24 METHODOLOGICAL, PSYCHOLOGICAL, PHILOSOPHICAL, STATISTICAL FOUNDATIONS back-test results and make rule evaluation difﬁcult.

…

That price change is given as the log of the ratio deﬁned as opening price of day+2 divided by the opening price on day+1, as shown in the following equation: Pos 0 × O+2 O+1 Log Where: POS 0 = Rule’s market position as of the close of day0 O+1 = Open S&P 500 on day+1 O+2 = Open S&P 500 on day+2 This equation does not show the detrended version of rule returns, as shown here: Pos 0 × Log O+2 O+1 – ALR Where: POS 0 = Rule’s market position as of the close of day0 O+1 = Open S&P 500 on day+1 O+2 = Open S&P 500 on day+2 ALR = Average Log Return over Back Test Look-ahead bias can also infect back-test results when a rule uses an input data series that is reported with a lag or that is subject to revision. For example, the back-test of a rule that uses mutual fund cash statistics,15 which is released to the public with a two-week delay, must take this lag into account by lagging signals to reﬂect the true availability of the data. None of the rules tested in this book use information reported with a lag or that is subject to revision. Objective Rules and Their Evaluation 31 Trading Costs Should trading costs be taken into account in rule back-tests? If the intent is to use the rule on a stand-alone basis for trading, the answer is clearly yes.

…

This is illustrated in Figure 6.4. In data mining, the back-test performance statistic plays a very different role than it does in single-rule back testing. In data mining, back- Cumulative Gains $ Expected Performance Observed Performance Back-Test +/Random Variation In-Sample Future Time FIGURE 6.4 Expected performance for single rule back test. 271 Data-Mining Bias: The Fool’s Gold of Objective TA tested performance serves as a selection criterion. That is to say, it is used to identify the best rule. The mean returns of all back-tested rules are compared and the one with the highest return is selected.

pages: 443 words: 51,804

Handbook of Modeling High-Frequency Data in Finance
by Frederi G. Viens , Maria C. Mariani and Ionut Florescu
Published 20 Dec 2011

Published 2012 by John Wiley & Sons, Inc. 421 422 Augmented log likelihood, 172 Autocorrelation, of GARCH ﬁltering, 202 Autocorrelation function (ACF), 177, 221 for minute data, 202–203 Automated trading platforms, 235 Automated trading systems, 63–64, 68 Autoregressive conditional duration (ACD) model, 27–28 Autoregressive conditionally heteroskedastic (ARCH) models, 272 Average daily volume (ADV), 34 classiﬁcation of equity based on, 45 Average estimator, 279 BAC data series, DFA and Hurst methods applied to, 155 Backtest, evaluating results of, 192 Backtest algorithm, 189 Backtest failure ratio, 192 Backtesting, 188–203 Backtest null hypothesis, 202 Backtest results, using GARCH, 204–205 Backtest result tables, 192–195, 199–200 Backtest variant, 195–196 Balanced capital structure, 59 Balanced scorecards (BSCs), 48, 52–53, 69. See also Board balanced scorecards (BSCs); BSC entries; Enterprise BSC; Executive BSC Ball solution, 391–399 Banach spaces, 349, 350, 351, 386, 387–388, 389 Bandwidth choices, 269 Barany, Ernest, xiii, 119, 327 Bartlett-type kernels, 261, 263 Base learner, 48 Bear Stearns crash, high-frequency data corresponding to, 121, 131–132 Bear Stearns crash week, high-frequency data from, 148–160 Beccar Varela, Maria Pia, xiii, 119, 327 Bernoulli LRT, 191.

…

Since the failure ratio of the backtests 5/216≈2.3% is much lower than 5%, we consider our model performance is satisfactory. TABLE 7.4 Backtest Results: Two Days Violation Ratio q Group 1 Group 2 12 Since 0.05 0.059 0.050 0.025 0.026 0.027 0.01 0.010 0.010 p-Value 0.005 0.005 0.007 0.05 0.076 0.975 0.025 0.809 0.596 0.01 0.827 0.827 0.005 0.864 0.277 the sum of i.i.d. Bernoulli r.v. is a binomial r.v., another alternative is a standard two-sided binomial test, as described by Casella and Berger (2002). 13 We reject the null hypothesis when the p-value is <0.05. 193 7.5 Backtesting TABLE 7.5 Backtest Results: Three Days Violation Ratio q Group 1 Group 2 Group 3 0.05 0.057 0.054 0.052 0.025 0.027 0.032 0.025 0.01 0.012 0.013 0.012 p-Value 0.005 0.008 0.009 0.007 0.05 0.260 0.532 0.807 0.025 0.663 0.143 0.921 0.01 0.513 0.239 0.513 0.005 0.175 0.091 0.313 TABLE 7.6 Backtest Results: Four Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 0.05 0.055 0.041 0.052 0.053 0.025 0.025 0.030 0.023 0.030 0.01 0.012 0.013 0.013 0.020 p-Value 0.005 0.004 0.006 0.005 0.008 0.05 0.490 0.202 0.801 0.690 0.025 0.942 0.378 0.728 0.378 0.05 0.514 0.722 0.630 0.849 0.400 0.025 0.828 0.804 0.485 0.804 0.048 0.01 0.624 0.426 0.426 0.006 0.005 0.730 0.574 0.902 0.170 TABLE 7.7 Backtest Results: Five Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 0.05 0.045 0.053 0.046 0.052 0.057 0.025 0.024 0.026 0.029 0.026 0.037 0.01 0.007 0.012 0.011 0.016 0.012 p-Value 0.005 0.005 0.008 0.008 0.011 0.005 0.01 0.317 0.612 0.876 0.136 0.612 0.005 0.913 0.293 0.293 0.059 0.913 TABLE 7.8 Backtest Results: Six Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 0.05 0.052 0.049 0.056 0.048 0.057 0.051 0.025 0.019 0.022 0.025 0.021 0.030 0.032 0.01 0.011 0.013 0.011 0.014 0.011 0.011 p-Value 0.005 0.006 0.008 0.008 0.011 0.006 0.006 0.05 0.785 0.927 0.529 0.782 0.421 0.927 0.025 0.318 0.649 0.949 0.470 0.422 0.298 0.01 0.783 0.514 0.783 0.310 0.783 0.783 0.005 0.645 0.336 0.336 0.061 0.645 0.645 194 CHAPTER 7 Risk Forecasting with Multiple Timescales TABLE 7.9 Backtest Results: Seven Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 0.05 0.043 0.044 0.041 0.044 0.039 0.054 0.059 0.025 0.024 0.017 0.033 0.024 0.024 0.020 0.030 0.01 0.011 0.013 0.013 0.013 0.019 0.011 0.006 p-Value 0.005 0.007 0.006 0.004 0.006 0.013 0.007 0.006 0.05 0.418 0.546 0.308 0.546 0.218 0.696 0.337 0.025 0.890 0.187 0.238 0.890 0.890 0.477 0.503 0.01 0.799 0.508 0.508 0.508 0.075 0.799 0.257 0.005 0.459 0.857 0.654 0.857 0.029 0.459 0.857 TABLE 7.10 Backtest Results: Eight Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 0.05 0.034 0.055 0.047 0.038 0.051 0.053 0.044 0.049 0.025 0.021 0.023 0.028 0.017 0.030 0.017 0.023 0.023 0.01 0.013 0.015 0.008 0.015 0.008 0.008 0.015 0.006 p-Value 0.005 0.006 0.006 0.004 0.006 0.004 0.006 0.008 0.004 0.05 0.089 0.618 0.733 0.218 0.933 0.770 0.576 0.899 0.025 0.586 0.811 0.728 0.235 0.528 0.235 0.811 0.811 0.05 0.158 0.649 0.492 0.096 0.822 0.511 0.824 0.824 0.355 0.025 0.127 0.875 0.647 0.631 0.875 0.877 0.451 0.875 0.245 0.01 0.570 0.325 0.732 0.325 0.732 0.732 0.325 0.394 0.005 0.689 0.689 0.809 0.689 0.809 0.689 0.331 0.809 TABLE 7.11 Backtest Results: Nine Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 0.05 0.036 0.045 0.043 0.033 0.048 0.057 0.052 0.052 0.040 0.025 0.014 0.024 0.029 0.021 0.024 0.026 0.031 0.024 0.017 0.01 0.007 0.012 0.010 0.010 0.007 0.014 0.017 0.014 0.010 p-Value 0.005 0.005 0.005 0.005 0.002 0.005 0.010 0.014 0.010 0.002 0.01 0.535 0.703 0.921 0.921 0.535 0.407 0.210 0.407 0.921 0.005 0.944 0.944 0.944 0.396 0.944 0.243 0.028 0.243 0.396 195 7.5 Backtesting TABLE 7.12 Backtest Results: 10 Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10 0.05 0.040 0.048 0.048 0.040 0.034 0.032 0.040 0.048 0.042 0.053 0.025 0.016 0.026 0.024 0.019 0.016 0.016 0.016 0.026 0.026 0.032 0.01 0.011 0.016 0.021 0.013 0.005 0.013 0.008 0.011 0.016 0.011 p-Value 0.005 0.008 0.008 0.016 0.008 0.005 0.008 0.008 0.005 0.008 0.005 0.05 0.340 0.831 0.831 0.340 0.141 0.082 0.340 0.831 0.483 0.797 0.025 0.224 0.858 0.881 0.398 0.224 0.224 0.224 0.858 0.858 0.420 0.01 0.910 0.290 0.058 0.548 0.312 0.548 0.676 0.910 0.290 0.910 0.005 0.456 0.456 0.017 0.456 0.937 0.456 0.456 0.937 0.456 0.937 TABLE 7.13 Backtest Results: 15 Days Violation Ratio q Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9 Group 10 Group 11 Group 12 Group 13 Group 14 Group 15 0.05 0.046 0.049 0.044 0.058 0.054 0.054 0.054 0.060 0.056 0.056 0.061 0.065 0.056 0.058 0.058 0.025 0.036 0.034 0.029 0.031 0.027 0.031 0.041 0.034 0.037 0.029 0.041 0.032 0.032 0.031 0.029 0.01 0.015 0.014 0.020 0.012 0.015 0.017 0.020 0.019 0.017 0.015 0.014 0.014 0.014 0.014 0.015 p-Value 0.005 0.010 0.010 0.010 0.007 0.012 0.012 0.007 0.007 0.012 0.012 0.009 0.007 0.009 0.012 0.010 0.05 0.645 0.940 0.512 0.395 0.627 0.627 0.627 0.303 0.504 0.504 0.227 0.119 0.504 0.395 0.395 0.025 0.118 0.184 0.553 0.399 0.735 0.399 0.024 0.184 0.072 0.553 0.024 0.277 0.277 0.399 0.553 0.01 0.230 0.405 0.026 0.652 0.230 0.121 0.026 0.058 0.121 0.230 0.405 0.405 0.405 0.405 0.230 0.005 0.117 0.117 0.117 0.557 0.044 0.044 0.557 0.557 0.044 0.044 0.274 0.557 0.274 0.044 0.117 Still, the high–low frequency method does have its limits.

…

As the time horizon stretches, we are ignoring more and more new information, leading to deteriorating GARCH performance. 7.5.5 A VARIANT BACKTEST: FORECASTING A ONE-DAY RETURN, n DAYS AHEAD The backtest above evaluates the accuracy of the simulated sum ni=1 Xt+i as a whole. To further examine each term inside that sum, we can slightly modify the previous backtest algorithm. 196 CHAPTER 7 Risk Forecasting with Multiple Timescales ALGORITHM 7.26 Alternative Backtest For t = C, C + n, C + 2n, . . . , C + kn, . . . 1. Apply GARCH ﬁlter to {Xt−C +1 , . . . , Xt }, and calibrate the skewed t distribution. α (Xt+n |Ft ), that is, VaR for day t + n, for α = 2.

pages: 504 words: 139,137

Efficiently Inefficient: How Smart Money Invests and Market Prices Are Determined
by Lasse Heje Pedersen
Published 12 Apr 2015

In summary, you could find trading strategies by getting an edge in trading and financing illiquid securities or by trading against demand pressures. 3.3. HOW TO BACKTEST A TRADING STRATEGY Once you have a trading idea, backtesting it can be a powerful tool. To backtest a trading strategy means to simulate how it would have done historically. Of course, historical performance does not necessarily predict future performance, but a backtest is very useful nevertheless. For instance, many trading ideas are simply born bad, and this can be discovered through a backtest. For instance, suppose you have a trading idea, simulate how it would have performed over the past 20 years, and find that the strategy would never have worked in the past.

…

Furthermore, some version will have worked the best in the past, perhaps just by chance, but, if this is by chance, it probably will not work well in the future, when you are actually trading on it. Or you tried the backtest because you heard someone made money on this trade, but, in this case, the backtest is biased to look good (your friend already told you!), even if this is by pure chance. These unavoidable biases mean that we should discount backtest returns and place more weight on realized returns. Furthermore, we should discount backtests more if they have more inputs and have been tweaked or optimized more. While unavoidable biases should simply affect how we should regard backtests, there are many avoidable biases that experienced traders and researchers fight hard to eliminate.

…

Given these expected transaction costs, we can adjust the backtest in the following simple way. Each time a trade takes place in our backtest, we compute the expected transaction cost and subtract this cost from the backtest returns. For instance, if we have a monthly portfolio rebalance rule, then each month of the backtest, we do the following: • Compute the return on the portfolio, • Compute the new security positions and the implied trades, • Compute the expected trading costs for every security and add them up, and • Subtract the total expected trading cost from the portfolio return. Adjusting a backtest for transaction costs is more important the higher the turnover of the trading strategy.

pages: 367 words: 97,136

Beyond Diversification: What Every Investor Needs to Know About Asset Allocation
by Sebastien Page
Published 4 Nov 2020

It’s finance 101, and again I suppose there’s nothing more practical than a good theory. In my backtests, when I replaced the economic growth model estimates with the sustainable growth rate for each country, the strategy’s performance jumped significantly. I was young, starting my career, and I was quite excited to see the strategy beat the pants off the MSCI World Index, with very limited look-ahead bias. (In hindsight, exceptional backtest performance is never that exceptional. A client once told me that he had never seen a backtest that didn’t work. He added that the only people who can consistently generate Sharpe ratios of 3.0 or above were quants running backtests, plus Bernie Madoff.

…

Then I applied this beta to recent flows. It worked very well: the first time I ran the backtest with the full forecast of income, growth, and flow-based valuation change, the results improved drastically. They got even better when I introduced the ROE component for growth. When I first saw these results, I stood up at my desk and danced like a prospector who’s just struck oil. Mark set up the project as a horse race between models. First, he asked me to run a backtest without any return forecast. To run a backtest without return forecasts, I set up the optimizer to solve for the risk-minimizing weights every month.

…

William Fallon, James Park, and Danny Yu, in their 2015 paper “Asset Allocation Implications of the Global Volatility Premium,” backtest volatility risk premium strategies across 11 equity markets, 10 commodities, 9 currencies, and 4 government bond markets. They find that “the volatility risk premium is sizable and significant, both statistically and economically.” TABLE 7.2 Prior Studies on the Volatility Risk Premium The same caveats apply as for the managed volatility studies—namely, that only backtests with good results tend to get published and that authors often ignore implementation shortfall between backtests and realized performance. Nonetheless, in practice, covered call writing has been shown to deliver good risk-adjusted performance, although perhaps not as high as 6–7% alpha across all market regimes.

Risk Management in Trading
by Davis Edwards
Published 10 Jul 2014

However, this does not indicate that the approach will work in the future. Backtesting and Trade Forensics 99 KEY CONCEPT: RISK AND RETURN Risky investments are likely to have periods where they are both more profitable and have greater losses than safer investments. As a result, if trading tests only look at profitability, the top-performing investment in any given historical period might be an extremely risky investment that just happened to get lucky. Historical Backtesting Historical backtesting is the process of using financial data for prior periods to develop and examine potential trading strategies. Inherent in the backtesting process is a bias to selecting high-risk investments.

…

(See Figure 6.9, VAR Backtest.) This data can be analyzed to see if the number of samples outside the VAR estimate matches the confidence level of the VAR calculation. For example, in the NYMEX WTI graphic, 5.58 percent of the days had losses greater than the VAR threshold for the day. Given the 2200 daily observations in the backtest, this indicates that losses are more common than the 5 percent of samples estimated by VAR. NYMEX WTI Crude Oil FIGURE 6.9 01 3 01 2 /2 28 6/ 28 /2 01 1 6/ 6/ 28 /2 01 0 /2 9 /2 28 6/ 6/ 28 00 00 6/ 28 /2 /2 0 28 6/ /2 28 6/ VAR Backtest 8 07 6 00 00 /2 6/ 28 6/ 28 /2 00 5 4 20.00% 15.00% 10.00% 5.00% 0.00% −5.00% −10.00% −15.00% −20.00% 171 Value-at-Risk THE MISUSE OF VAR A major problem with VAR is that VAR gets used for multiple purposes.

…

The focus of the book is how traders working at hedge funds or on investment bank proprietary trading desks use risk management techniques to improve their profitability and keep themselves in business. However, these techniques can be applied to almost any trading or investment group. This book focuses on six major activities that are part of managing trading businesses. 1. Backtesting and Trade Forensics. Backtesting is a disciplined approach to testing trading ideas before making bets with actual money. Trade forensics 1 2 RISK MANAGEMENT IN TRADING 2. 3. 4. 5. 6. is a post‐mortem analysis that identifies how well a trade is tracking pre‐ trade predictions and if markets have changed since the trade was initiated.

pages: 327 words: 91,351

Traders at Work: How the World's Most Successful Traders Make Their Living in the Markets
by Tim Bourquin and Nicholas Mango
Published 26 Dec 2012

It sounds like you came across trend-following strategies that work for you, and they have continued to do well for years. Do you think that’s the case with most backtested strategies? German: Whether or not a backtested strategy does well over the long term depends on the way in which that particular strategy was backtested. A backtested strategy might look great on paper but not make any money in real life. That said, I will not trade anything that hasn’t been tested. Maybe I’m just wired this way now, because I have been backtesting strategies for so long, and it’s kind of engrained in me, but when somebody says they are doing this or that in the market, I always ask them, “How do you know your strategy works?

…

With regards to breakout strategies, when a market breaks out of a range—and it could be a five-day range or a five-hundred-day range—you enter your position in the direction of that breakout, and you fine-tune your stop based on your testing. Bourquin: Are you constantly backtesting markets to find where certain trends work best? Or have you backtested trades in the past so that you now know what approach works best in a given market and can apply it going forward? German: Initially, trend following involves a lot of backtesting and thousands and thousands of tests, including millions of iterations on all kinds of different markets, with all kinds of different trend-following strategies and approaches to stops and profit targets. There is an initial period where you are backtesting for ten hours a day, but then you get into a rhythm where you determine what works, based on your backtesting.

…

There is an initial period where you are backtesting for ten hours a day, but then you get into a rhythm where you determine what works, based on your backtesting. I did a great deal of backtesting over a couple of years, which solidified the markets that I wanted to trade, the program that I wanted to follow, and what does and does not work for me. Now, every time we experience a drawdown or every time I want to question or test myself, I will do some backtesting. But at the end of the day, I always go back to my original set of tests, and that is what I have been trading off of for years. Bourquin: Have you had to change your strategy at all? It sounds like you came across trend-following strategies that work for you, and they have continued to do well for years.

pages: 354 words: 26,550

High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems
by Irene Aldridge
Published 1 Dec 2009

Care should be taken to avoid “overfitting”—using the same data sample in repeated testing of the model. CHAPTER 16 Implementing High-Frequency Trading Systems nce high-frequency trading models have been identified, the models are back-tested to ensure their viability. The back-testing software should be a “paper”-based prototype of the eventual live system. The same code should be used in both, and the back-testing engine should run on tick-by-tick data to reenact past market conditions. The main functionality code from the back-testing modules should then be reused in the live system. To ensure statistically significant inferences, the model “training” period T should be sufficiently large; according to the central limit theorem (CLT), 30 observations is the bare minimum for any statistical significance, and 200 observations is considered a reasonable number.

…

Given strong seasonality in intra-day data (recurrent price and volatility changes at specific times throughout the day), benchmark high-frequency models are backtested on several years of tick-by-tick data. The main difference between the live trading model and the back-test model should be the origin of the quote data; the back-test system includes a historical quote-streaming module that reads historical tick data from archives and feeds it sequentially to the module that has the main functionality. In the live trading system, a different quote module receives real-time tick data originating at the broker-dealers. Except for differences in receiving quotes, both live and back-test systems should be identical; they can be built simultaneously and, ideally, can use the same code samples for core functionality.

…

Model Development The development of a high-frequency trading business begins with the development of the econometric models that document persistent relationships among securities. These relationships are then tested on lengthy spans of tick-by-tick data to verify the forecasting validity in various market situations. This process of model verification is referred to as “backtesting.” Standard back-testing practices require that the tests be run on data of at least two years in duration. The typical modeling process is illustrated in Figure 3.4. System Implementation The models are often built in computer languages such as MatLab that provide a wide range of modeling tools but may not be suited perfectly for high-speed applications.

pages: 263 words: 75,455

Quantitative Value: A Practitioner's Guide to Automating Intelligent Investment and Eliminating Behavioral Errors
by Wesley R. Gray and Tobias E. Carlisle
Published 29 Nov 2012

Companies often restate financial statements after the fact, and this can introduce another form of look-ahead bias that can have a huge impact on back-tested results. Marcus Bogue and Morris Bailey in their white paper, “The Advantages of Using as First Reported Data with Current Compustat Data for Historical Research,”16 highlight how restated financial statements impact back-test results for a simple price-to earnings ratio strategy. If the back-test fails to account for the difference in financial statement data as the data are first reported and then as they are subsequently restated, the back-test results vary dramatically. For example, from June 1987 through June 2001, failing to account for look-ahead bias caused by restatement of financial results led to an overstatement of returns achievable with the price-to-earnings ratio strategy by an incredible 28 percent.

…

In this chapter, we discuss our philosophy for conducting investment simulations, and survey the potential pitfalls in interpreting back-test results. We cast a suspicious eye on back-tested, and real, historical results, closely scrutinizing the steps we can take to ensure that results are genuine, and replicable. In Chapter 11, we study the best way to combine the research we've already considered into a cohesive strategy. We examine the Magic Formula and the F_SCORE to see if we can find a better structure for our valuation model. Our process leads us to identify some potential structural issues with the Magic Formula. In Chapter 12, the final chapter, we back-test the quantitative value model we created in Chapter 11.

…

There we look at the impact of buybacks, insider purchases, short selling, and buying and selling from institutional investment managers like activists and other fund managers. Finally, in Part Six we build and test our quantitative value model. We study the best way to combine the research we've considered into a cohesive strategy, and then back-test the resulting quantitative value model. CHAPTER 1 The Paradox of Dumb Money “As they say in poker, ‘If you've been in the game 30 minutes and you don't know who the patsy is, you're the patsy.'” —Warren Buffett (1987) In the summer of 1968, Ed Thorp, a young math professor at the University of California, Irvine (UCI), and author of Beat the Market: A Scientific Stock Market System (1967), accepted an invitation to spend the afternoon playing bridge with Warren Buffett, the not-yet-famous “value” investor.

pages: 241 words: 81,805

The Rise of Carry: The Dangerous Consequences of Volatility Suppression and the New Financial Order of Decaying Growth and Recurring Crisis
by Tim Lee , Jamie Lee and Kevin Coldiron
Published 13 Dec 2019

Since we are focused on the practical consequences of the growth of carry, our study of historical returns aims to be as realistic as possible while recognizing that no backtest can accurately replicate history. We therefore base our analysis on a backtest of a hypothetical currency carry portfolio built using simple but realistic rules that could easily be replicated in practice. Details of how we constructed our backtest are provided in the box (“Currency Carry Backtest”) at the end of this chapter. Our backtest allows us to highlight the main features of the currency carry trade, features that other types of carry can be expected to share.

…

The opportunity for profit in developed country currency carry has shrunk considerably over the last 40 years and is at an all-time low in recent years. We can see this shrinking opportunity set directly by looking at the actual carry a backtest portfolio earns. The carry is simply the interest rate the portfolio earns on its long positions less the implied cost of borrowing on the short positions. In Figure 4.6 we present this for a portfolio that goes long $1 and short $1 for each dollar of capital. In contrast, our standard backtest employs five times this exposure, so the carry would be multiplied by five. But we choose this presentation to keep the units on a magnitude that is familiar and also more directly comparable to the preceding graph on the raw difference between maximum and minimum interest rates.

…

TABLE 4.2 Average Size of Carry Portfolio’s Biggest Long Positions, 2014–2018 Country Average Long Position Turkey 11.8% Brazil 11.4% Russia 10.9% South Africa 9.6% India 9.5% Mexico 8.0% Indonesia 7.2% China 6.3% New Zealand 5.4% Malaysia 5.4% Australia 4.2% Source of data: Datastream, Global Financial Data, authors’ calculations In terms of developed markets, the Australian and New Zealand dollars stick out as consistent recipients of carry trade flows, which is also consistent with BIS economic data. Our backtest that only uses developed currencies allocates over half its long positions in the last decade to these two countries. Even in a global carry strategy that includes emerging currencies, these two countries retain their positions in the long portfolio. 64 THE RISE OF CARRY Currency Carry History Lessons As discussed in the previous chapters and explored further in subsequent chapters, carry trades are, by their nature, levered. Our backtest history tells us that purely as an investment strategy, currency carry needs to employ substantial leverage—total gross positions of 10 times capital—in order to generate absolute levels of returns that are competitive with conventional asset classes.

pages: 483 words: 141,836

Red-Blooded Risk: The Secret History of Wall Street
by Aaron Brown and Eric Kim
Published 10 Oct 2011

Nevertheless, historical simulation VaR has become the most popular VaR number for risk reporting and regulatory purposes. Why? Because it’s easy to compute and its objective. It never surprises you; in fact, it’s always pretty close to yesterday’s value. The fact that it can’t pass a back-test doesn’t matter to people who never look at back-tests. The fact that it is actively misleading, telling you it’s safe when it’s dangerous and telling you it’s dangerous when it’s safe, doesn’t matter to people who only report and regulate. That only matters to people who manage risk. One fix that might occur to you is to set the VaR halfway between the fifth and sixth worst losses instead of between the 10th and 11th worst.

…

There are many problems with this approach, the biggest theoretical one is the average prediction made assuming an average value is exactly right, is exactly wrong. The biggest practical one is it never back-tests well. Moreover, variance-covariance VaR tells you even less about the tails than other VaRs. However, it was the form in which JPMorgan introduced VaR to the world. Some people still think of it as the definition of VaR. JPMorgan needed it to produce a report within 15 minutes of market close, using 1990-era technology and data systems. Variance-covariance was the only practical option. For Basel II, however, many flavors of VaR were easily available and we should have insisted on one that can pass back-test. But the most momentous decision, which seemed innocuous at the time, was to promise banks capital relief for spending all the money to create Basel II systems.

…

We ended up stealing methods from the people who set sports betting point spreads, and adding stuff we made up as we went along. We had to delve deeply into the back office and study methods developed by controllers and auditors. Only after years of intensive, cooperative work did we develop VaRs that could pass rigorous statistical back-tests, and on which we were willing to bet with traders. The only way you got VaR accepted on the trading floor in the early 1990s was to bet; you can imagine what traders think of a risk manager who tells them how to run their billion-dollar portfolios but won’t risk $10,000 of his own money on his analysis.

Alpha Trader
by Brent Donnelly
Published 11 May 2021

I would say Superforecasting is one of the books that has most influenced my quantitative thinking style (along with Fooled by Randomness by Nassim Taleb, Risk by Dan Gardner, and Thinking in Bets by Annie Duke). A full list of book recommendations appears in Appendix A. Backtest your ideas. If you notice that stocks tend to fall every Monday and rally every Tuesday, backtest it. If it feels like oil rallies every day from 2PM to 3PM, check it out. Is it true or false? Whenever a hypothesis appears in your head, dig in and figure out if it’s true. Playing with ideas and data like this can inspire new ideas and a better understanding of the inner workings of the markets you trade. If you don’t know how to backtest, just open Excel and start messing around. This is the kind of thing you can teach yourself with a bit of effort.

…

Chart courtesy of Refinitiv I got set on the trade the day after payrolls (at 107.75) and USDJPY went pretty much straight down. Backtesting is an entire field of quantitative research and there is considerable complexity involved. I am not a computational finance or applied math expert, so I like to keep things as simple as possible. I come up with a theory, and then I test it. I don’t tweak the parameters or snoop around until I find something useful. I just go in, test, and get out. While a full course in backtesting is beyond the scope of this book, you don’t need a degree in financial engineering to conduct basic backtests of patterns you identify and ideas you come up with.

…

The simplest way to be curious is to simply ask “Why?” as much as possible. Back in the day, when everyone knew that apples fell from trees, Newton was the one to ask: “Why don’t apples fall sideways? Why don’t they fall upwards?” When you see a phenomenon in the markets, ask why. This is particularly true when backtesting. When you discover a pattern that backtests successfully, can you explain why it works? If you can, there is a much greater probability that what you discovered will work out of sample. Much of what you find sifting through data is just data snooping or data mining. But when you find patterns and can explain them logically, you have something much more powerful.

Investing Amid Low Expected Returns: Making the Most When Markets Offer the Least
by Antti Ilmanen
Published 24 Feb 2022

But, none of this changes the sentiment that strategies getting very cheap or very expensive over the period studied do matter a ton in the real world over time horizons investors care about. We have often used the physics-envy term “time dilation” to refer to how long real life can seem while living it versus how short it can seem when checking out a backtest. It's easy to look at a good backtest in an intuitive robust strategy and examine its three- to five-year painful periods and think, “Of course I'd stick with it, it makes economic sense and look at the whole history!” But, I'm guessing, to none of our readers' surprise, it's a little harder to do live and real-time!

…

Both concerns may warrant discounting the historical performance, but for quite different reasons. The overfitting argument implies that the opportunity identified in backtests (or selections among many live strategies) was never as strong ex ante as was reported ex post. The overcrowding argument implies that the opportunity really was there, but the world has changed, and the greater awareness and greater popularity of a strategy implies lower rewards going ahead. Choosing how much to discount historical backtests – to subtract the impact of trading costs and either of the overfitting or “the world has changed” effects – inevitably involves judgment.

…

Trend Cumulative Performance, Jan 1985–Mar 2020 Source: Data from AQR. The hypothetical Put strategy is a backtest which involves buying a 5% out-of-the-money one-month put on the S&P500 index (pre-1996 on the S&P100) at mid-month and rebalancing into a new put at expiry. Put returns are expressed as a percentage of the underlying index NAV, gross of trading costs and fees. For comparability, the series is scaled to 10% volatility based on the 6% volatility of the unlevered return over the full sample, implying a leverage of 1.67. The hypothetical Trend return is a backtest, gross of fees, net of estimated transaction costs. The strategy applies trend following at 1-, 3- and 12-month windows in four asset classes and targets overall portfolio volatility of 10%.

Unknown Market Wizards: The Best Traders You've Never Heard Of
by Jack D. Schwager
Published 2 Nov 2020

Oh, so you didn’t eliminate stops, you eliminated intraday stops. That’s right. I found that by using intraday stops, I was getting stopped out a lot on meaningless intraday volatility. Were there any other significant changes that resulted from your initial backtesting efforts? The most significant impact was that once I got involved in building backtesting software, I became obsessed with it. At heart, I am more of a software developer than a trader. By mid-1999, I had become skeptical that there was any connection between how nice a chart pattern looked and the probability of a successful trade. I wanted to delegate chart evaluation to my software.

…

It’s wise to include an on/off switch (e.g., an equity curve moving below its moving average) in each strategy, even if it reduces profits in a backtest. This rule can significantly limit losses if a system stops working, as can and will happen. The more strategies you run, the easier it is emotionally to turn one off. In this sense, this rule reinforces the importance of rule #1. The tail-risk in a mean-reversion strategy is more likely to come from a cluster of medium-sized losing trades rather than from the potential for a huge loss in an individual trade. Allow for the possibility that the serial correlation of losses is probably understated in any backtest. It is possible to succeed quickly, continue to do well for 15 years, and then have a near-career-ending drawdown.

…

Did he give you any specific advice? He told me it was essential to go as far back as I could to test my views. Dalio is a master of history. He said that people get blindsided by something that has happened before because they overweight recent history and their own experience and don’t go back far enough to backtest their views. He helped me understand that to get conviction, you have to look at history well beyond your own experience. So now, when you analyze markets, how far back do you go? As far as I can. Which means what? At the moment, about 100 years—but, ideally, I would like to go back well beyond that.

High-Frequency Trading
by David Easley , Marcos López de Prado and Maureen O'Hara
Published 28 Sep 2013

We next show how bad information leakage increases execution shortfall and introduce the BadMax approach for testing whether algos leak information to predators. In the BadMax approach, we pretend to be BadMax, a fictitious predator. As BadMax, we use historical data to back-test whether BadMax can construct profitable predatory strategies. We next describe the historical data we use in our back-tests and estimate the BadMax gross and net Alpha for several BadMax predatory strategies. In our back-tests we assume that BadMax can somehow 159 i i i i i i “Easley” — 2013/10/8 — 11:31 — page 160 — #180 i i HIGH-FREQUENCY TRADING Figure 8.1 Good and bad information leakage (hypothetical example) Price over AlphaMax execution horizon (bp) GoodMax sells XYZ 12 10 BadMax buys XYZ 0 Open 8 6 4 AlphaMaxbuy executions XYZ price if AlphaMax did not buy drive XYZ price up 2 0 10h00 0 11h00 0 12h00 0 13h00 0 14h00 0 15h00 AlphaMax buys 10K shares AlphaMax buys 10K shares AlphaMax buys 10K shares AlphaMax buys 10K shares AlphaMax buys 10K shares AlphaMax buys 10K shares 0 Close First Second Third Fourth Fifth Sixth tranche tranche tranche tranche tranche tranche identify GSET buy and sell algo executions.

…

We repeated the calculation for extreme clusters with more than 100 marketable GSET algo executions; in this case, GSET algo prints account for only 14% of all Tape prints. Even large clusters of marketable GSET algo executions, therefore, leave almost no footprint on the Tape. BadMax cannot extract GSET algo Alpha by analysing the Tape. CONCLUSION Based on the predatory strategies we have back-tested so far, GSET algo executions do not leak information that predators can profitably exploit by trading along. Our back-tests show the following. • Non-marketable and peg mid algo executions are associated with negative Alpha and therefore do not leak information that predators can exploit. • Marketable algo executions are associated with low positive Alpha but this Alpha does not cover the predators’ roundtrip cost; marketable executions, therefore, also do not leak information that predators can exploit. • Large clusters of marketable algo executions are associated with high positive Alpha, but, because these clusters are expensive to identify, predators can capture little if any Alpha.

…

In practice, therefore, it is futile to try to test whether an algo leaks information by comparing algo performance. We developed, instead, the BadMax approach. THE BADMAX APPROACH AND DATA SAMPLE In order to test whether algo executions leak information, we pretend to be a fictitious predator, BadMax. As BadMax, we use historical data on GSET algo executions to back-test the profitability of different information leakage scenarios. In one test, for example, we assume that BadMax can identify marketable GSET algo buy executions and test whether BadMax can use this information to generate profitable buy signals. Our tests assume BadMax has more information on GSET algo executions than any real-world predator is ever likely to have.

pages: 447 words: 104,258

Mathematics of the Financial Markets: Financial Instruments and Derivatives Modelling, Valuation and Risk Issues
by Alain Ruttiens
Published 24 Apr 2013

If the fund strategy is stable enough over time, the VaR calculation can be more accurate than through the traded instruments, first by avoiding the correlations problem, and second, if the composition of the fund portfolio is often modified. Backtesting of the VaR VaR methods presenting several weaknesses – starting with the adequate selection among several VaR methodologies – a VaR estimate needs to be tested a posteriori (“backtested”), to check to what extent it fits with actually observed losses larger than the VaR amount. As said by A. Brown11, “VaR is only as good as its backtest. When someone shows me a VaR number, I don't ask how it is computed. I ask to see the backtest”. The simple way to check it consists in counting the number N of times a portfolio presents losses that exceed the VaR number on a series of n successive VaR calculations.

…

The simple way to check it consists in counting the number N of times a portfolio presents losses that exceed the VaR number on a series of n successive VaR calculations. Depending on N/n, – called the “failure rate” – being higher or lower than the confidence level c associated with the VaR measures, the used VaR model is over- or underestimating the risk (the equality between N/n and c being obviously very unlikely). The most popular backtest is the Kupiec's one, also called “POF (for Proportion of failures) test”. In this test, the losses exceeding the VaR number are considered to be independently and identically distributed, so that N follows a binomial distribution f(N) (that a loss may exceed or not the VaR number). For a confidence level c, the corresponding frequency of losses p is, repeating Eq. 14.5, (14.5) so that the failure rate N/n could be used as an unbiased measure of p, that would converge to 1 − c with n growing. f(N) is therefore described by the binomial distribution where CnN denotes the number of possible combinations of N failures and (1 − N) “non-failures” on a total of n events.

…

LEDOIT, “Gain, loss and asset pricing”, The Journal of Political Economy, Jan 2000, pp. 144–172. 7 This section is partly inspired from Philippe JORION, Financial Risk Manager Handbook, 5th ed., 2009, John Wiley & Sons, Inc., Hoboken, and Moorad CHOUDHRY, An Introduction to Value-at-Risk, 4th ed., 2006, John Wiley & Sons, Ltd, Chichester. 8 To make a more precise calculation, the width of the bins should be narrower than 0.5%, as used here. 9 The 2510 returns used for the example present a kurtosis of 7.81 and a skewness of −0.10. 10 In the initial basic example, the only risk factor was the price change of the exposure in S&P 500. 11 A. BROWN, Private Profits and Socialized Risk – Counterpoint: Capital Inadequacy, Global Association of Risk Professionals, June/July 08 issue. Cited by O. NIEPPOLA in his masters Thesis Backtesting Value-At-Risk Models, Helsinki School of Economics, 2008. 12 See any book of statistics. This ratio is a log ratio of the likelihood that p = , divided by the likelihood that p is not = . To verify this, one must use the values of a χ2 distribution, if p is asymptotically Gaussian, which is the case of a binomial distribution. 13 See Peter F.

pages: 464 words: 117,495

The New Trading for a Living: Psychology, Discipline, Trading Tools and Systems, Risk Control, Trade Management
by Alexander Elder
Published 28 Sep 2014

System Testing, Paper Trading, and the Three Key Demands for Every Trade Before trading real money with a system, you need to test it, whether you developed it yourself or bought it from a vendor. This can be done in one of two ways. One is backtesting: apply your system's rules to a stretch of historical data, usually several years' worth. The other is forward-testing: trade small positions with real money. Serious traders begin with backtesting, and if its results look good, switch to forward-testing; if that works well, they gradually increase position size. Looking at printouts of historical results is a nice start, but don't let good numbers lull you into a false sense of security.

…

Your drawdown is deepening, and then the system flashes a new signal. Will you put on the next trade? Suddenly, an impressive printout looks like a very thin reed on which to hang the future of your account. There is a cottage industry of programmers who back-test systems for a fee. Some traders, too suspicious to disclose their “sure-fire methods,” spend months learning to use testing software. In the end, only one kind of backtesting prepares you to trade—manual testing. It is slow, time-consuming, and cannot be automated, but it's the only method that comes close to modeling real decision making. It consists of going through historical data one day at a time, scrupulously writing down your trading signals for the day ahead, and then clicking one bar forward and recording new signals and trades for the next day.

…

Many have a fantasy that successful traders have some secret knowledge. That fantasy helps support a lively market in advisory services and ready-made trading systems. A demoralized trader may whip out his credit card to buy access to “trading secrets.” He may send money to a charlatan for a $3,000 “can't miss,” backtested, computerized trading system. When that system self-destructs, he'll pull out his almost-maxed-out credit card again for a “scientific manual” that explains how he can stop losing and begin winning by contemplating the moon, the stars, or even Uranus. At an investment club we used to have in New York, I often ran into a famous financial astrologer.

pages: 120 words: 39,637

The Little Book That Still Beats the Market
by Joel Greenblatt
Published 2 Jan 2010

After I wrote the original edition, a number of Wall Street firms did conduct some research into this question (showing that the formula worked in pretty much all foreign markets tested), but we did not conduct any of our own backtests, for two reasons. First, much of the available historical stock market data from outside the United States is seriously flawed, and backtest results would not be reliable. It is helpful to know, however, that most historical studies over the last several decades involving classic (and less problematic to test) value characteristics, such as low price to earnings, low price to book value, and low price to sales have proved equally effective in both the United States and international markets.

…

Despite all its flaws, the formula certainly seems to have worked well over the long-term (fortunately, I received many nice e-mails about this, too). But over the last 10 years, the results from our test of roughly the largest 1,000 companies in the United States (with market capitalizations over $1 billion) tell an interesting tale. This is one of those rare 10-year periods over which the S&P 500 index was actually down. According to our backtests, on the other hand, the formula managed to earn 255 percent during this same period (more than tripling our money!). That’s a 13.5 percent annualized return during a 10-year period when the S&P index was actually down 0.9 percent per year. TABLE A.1 Updated Magic Formula Results Through 2009 But here’s the thing.

…

It is not easy to find an effective short-term hedging strategy for our favorite magic formula stocks. As a result, most of the benefits of the formula will continue to go to the much smaller group of investors who can maintain a true long-term horizon. One additional characteristic of the magic formula strategy is not necessarily good or bad. However, based on our updated backtests, it’s probably helpful to keep this one in mind. Over the last 22 years, when comparing the performance of the magic formula portfolios during up months for the S&P 500 and down months for the same index, it turns out that much of the outperformance of our portfolios comes during the up months. On average during this 22-year period, the magic formula portfolios “captured” 95 percent of the S&P 500’s performance during down months and 140 percent of its performance during up months.

The Unusual Billionaires
by Saurabh Mukherjea
Published 16 Aug 2016

Amongst test cricketers, Rahul Dravid was the first among equals. As Edward Smith, the former England and Kent batsman who batted alongside Dravid in county cricket matches says, Exhibit 6: Back-testing results of completed six iterations of the Coffee Can Portfolio (i.e. these iterations have run their complete course of ten years) Source: Bloomberg, Capitaline, Ambit Capital research. Note: Portfolio at start denotes an equal allocation of Rs 100 for the stocks qualifying to be in the CCP for that year. Exhibit 7: Back-testing results of incomplete ten iterations of the Coffee Can Portfolio (i.e. these iterations have not run their complete course of ten years) Source: Bloomberg, Capitaline, Ambit Capital research.

…

Thus I have ten CCPs which are incomplete in the sense that they have not yet run for the full duration of ten years. For these uncompleted CCPs, I have used a cut-off date of 5 April 2016 for measuring price performance. The results can be summarized in one sentence: Each of the sixteen CCPs (six with complete price performance back-testing and ten partially complete back-testing) has outperformed the benchmark large-cap index in India, the Sensex. In fact, as can be seen in Exhibit 141, the outperformance of the CCP to the Sensex is almost always in excess of 4 percentage points per annum. (Note: in financial markets, this outperformance relative to the benchmark is called alpha.)

…

The maximum drawdown, which is the largest single drop from the peak to the bottom in the value of a stock, for both the portfolios in this period was also less than the maximum drawdown for the Sensex. Exhibit 141: Back-testing results of completed six iterations of the Coffee Can Portfolio (i.e these iterations have run their complete course of ten years of price-performance check) Source: Bloomberg, Capitaline, Ambit Capital research. Note: *Portfolio at start denotes an equal allocation of Rs 100 for the stocks qualifying to be in the CCP for that year. Exhibit 142: Back-testing results of incomplete ten iterations of the Coffee Can Portfolio (i.e., these iterations have not run their complete course of ten years of price-performance check) Source: Bloomberg, Capitaline, Ambit Capital research.

The Permanent Portfolio
by Craig Rowland and J. M. Lawson
Published 27 Aug 2012

Chapter 3 Permanent Portfolio Performance Testing Theories When discussing any investment strategy, the question of how it has performed in the past always comes up. The process of looking at historical data to determine how a particular investment strategy may perform going forward is called “backtesting.” It's a good idea, however, to be cautious about drawing any firm conclusions about an investment strategy based solely on backtesting data. A better approach is to use backtesting as a tool to prove or disprove general ideas about a strategy, rather than to mechanically project past performance into the future. Backtesting can tell an investor how well an investment strategy's theories have worked in practice during historical periods. Thus, if something failed to work in the past there may be good reason to believe that it won't work in the future either.

…

Thus, if something failed to work in the past there may be good reason to believe that it won't work in the future either. Repeating mistakes and leaving to faith something that has failed once before is a bad (and likely expensive) strategy. Backtesting then can be an excellent tool to disprove theories or to provide a tentative validation of the way a theory would have worked in practice in the past. This approach is, however, quite different from simply noting that an investment strategy has gone up every year for the past 10 years and using that observation as evidence that the strategy will go up for the next 10 years.

…

In other words, at this point we have more than 30 years of empirical data to review to determine how the portfolio's theories have performed under real-world conditions. While this chapter provides performance data for the Permanent Portfolio, it also provides a framework for assessing other portfolio strategies as well. Two important questions that backtesting data can help to answer are: How do investing strategies fail and what actually causes them to fail? So how has the Permanent Portfolio done compared to other investment strategies? What sets it apart? Let's take a look. Growth, No Large Losses, and Real Returns: The Holy Trinity According to historical data, the Permanent Portfolio strategy has provided investment returns of 9 to 10 percent a year for the past 40 years.1 The worst loss in any year was around −5 percent back in 1981.

The Intelligent Asset Allocator: How to Build Your Portfolio to Maximize Returns and Minimize Risk
by William J. Bernstein
Published 12 Oct 2000

For example, a “simpleton’s portfolio” consisting of one quarter each U.S. large stocks, U.S. small stocks, foreign stocks, and U.S. high-quality bonds had a higher return, with much lower risk, than large U.S. stocks alone (represented by the S&P 500 index). The S&P 500, in turn, performed better than 75% of professional money managers over the same period. I was fascinated by the T. Rowe Price data; here was a simple tool for ascertaining historical asset allocation performance—collect data on the prior performance of various asset classes, and “backtest” returns and risks. To my disappointment, I could find no readily available software which accomplished this; I would have to write my own spreadsheet files. I began to buy, beg, steal or borrow data on a wide variety of assets over several different historical epochs and build portfolio models going back as far as 1926.

…

Clearly, at this level the use of 5-year notes is superior to the other two bond choices; over most of its extent it lies above the other two curves, indicating that for each degree of risk the 5-year notes and stock mix yields more return. Only at low risk levels is the use of T-bills desirable. Portfolio simulations with other databases using both backtesting and another technique called mean-variance Figure 4-2. Stock/bond mixes, 1926–1998. The Behavior of Real-World Portfolios 45 analysis also suggest the superiority of short-term bonds. On occasion it may be advantageous to use long-term bonds or T-bills in small amounts. In general, however, you will not go far wrong by sticking to bond maturities of six months to five years for the risk-diluting portion of your portfolio.

…

Who knows whether foreign or domestic stocks will have the higher return The Behavior of Real-World Portfolios 53 over the next 20, 30, or even 50 years? However, it seems highly likely that a 50/50 mix will not be too far from the best foreignversus-domestic allocation. The real purpose of portfolio backtesting, mean-variance analysis, or any other kind of portfolio analysis is not to find the “best” asset mix. Rather, it is to find a portfolio mix that will not be too far off the mark under a wide variety of circumstances. Small Stocks versus Large Stocks It’s important to realize how large and small stocks behave relative to each other.

pages: 320 words: 33,385

Market Risk Analysis, Quantitative Methods in Finance
by Carol Alexander
Published 2 Jan 2007

For instance, in Section II.5.5.3 we describe a pairs trade between the volatility index futures that have recently started trading on the CBOE and Eurex. Whenever a regression model is used to develop a trading strategy it is very important to backtest the model. Backtesting – which is termed out-of -sample testing or post-sample prediction by academics – is particularly important when considerable sums of money are placed on the output of a regression model. It is absolutely imperative to put the model through the rigorous testing procedure using a long period of historical data. A simple backtest proceeds as follows:28 1. Estimate the regression model on an historical sample of data on the variables, saving subsequent historical data for testing the model. 2.

…

For trading strategies we may look for strategies that produce a high Sharpe ratio, or that maximize some other risk adjusted performance measure, as outlined in Section I.6.5. But for a pure hedging strategy we should seek a hedge ratio that reduces risk alone, for instance one that minimizes the variance of the P&L. 28 This is an example of the general backtests described in Section II.8.5. 184 Quantitative Methods in Finance I.4.7 SUMMARY AND CONCLUSIONS This chapter has laid the foundations of regression analysis in the simple framework of linear models. We focus on the ordinary least squares (OLS) estimation criterion since this is optimal under fairly general circumstances when we have large samples of data, as is often the case in market risk analysis.

…

Statistical Tables 274 Statistical Tables Statistical Tables 275 276 Statistical Tables Statistical Tables 277 Index Abnormal return, CAPM 253 Absolute return 58 Absolute risk tolerance 231 Absolute value function 6 Active return 92, 256 Active risk 256 Alternative hypothesis 124, 151 American option 1, 215–16 Amex case study 144–6, 153–5 Amex Oil index 162–3, 169–70, 174 Analysis of variance (ANOVA) Amex case study 154 BHP Billiton Ltd case study 164–5 matrix notation 159–60 regression 143–4, 149–50 Analytic solution 185 Anderson–Darling test 128–9 ANOVA (analysis of variance) Amex case study 154 BHP Billiton Ltd case study 164–5 matrix notation 159–60 regression 143–4, 149–50 Appraisal ratio 257 Approximate confidence interval 122 Approximations delta–gamma 2, 34 delta–gamma–vega 34 duration–convexity 2–3, 34 finite difference 206–10, 223 Taylor expansion 31–4, 36 Arbitrage no arbitrage 2, 179–80, 211–13 pricing theory 257 statistical strategy 182–3 Arithmetic Brownian motion 22, 136, 138–9 Arrival rate, Poisson distribution 87–9 Ask price 2 Asset management, global 225 Asset prices binomial theorem 85–7 lognormal distribution 213–14 pricing theory 179–80, 250–55 regression 179–80 stochastic process 137–8 Assets, tradable 1 Asymptotic mean integrated square error 107 Asymptotic properties of OLS estimators 156 Autocorrelation 175–9, 184, 259–62 Autocorrelation adjusted Sharpe ratio 259–62 Autoregression 135 Auxiliary regression 177 Backtesting 183 Backward difference operator 19 Bandwidth, kernel 106–7 Bank 225 Barra model 181 Basic calculus 1–36 Basis splines 200 Bayesian probability 72–3 Bermudan option 1 Bernoulli trial 85–6 Best fit of function to data 201 Best fit line 145 Best linear unbiased estimator (BLUE) 157, 175 280 Index Beta values CAPM 252–3 diversifiable risk 181 OLS estimation 147–8, 156, 160–1, 183–4 regression 156 Bid–ask spread 2 Bid price 2 Bilinear interpolation 193–5 Binomial distribution 85–7, 213 Binomial lattices 186, 210–16, 223 American option 215–16 European option 212–13 lognormal asset price distribution 213–14 no arbitrage 211–12 risk neutral valuation 211–12 Bisection method, iteration 187–8 Bivariate distribution 108–9, 116–17, 148 Bivariate normal distribution 116–17, 148 Bivariate normal mixture distribution 116–17 Black–Scholes–Merton pricing model asset price 137–9 European option 2, 213, 215–16 lognormal distribution 94 numerical method 185 Taylor expansion 2–3 BLUE (best linear unbiased estimator) 157, 175 Bonds 1–2, 37, 191 Bootstrap 218 Brownian motion 136 arithmetic 22, 136, 139 geometric 21–2, 134, 138, 212, 213–14, 218–19 Calculus 1–36 differentiation 10–15 equations and roots 3–9 financial returns 16–26 functions 3–9, 26–31 graphs 3–9 integration 15–16 Taylor expansion 31–4, 36 Calibration 201 Call option 1, 6, 212–13 Capital allocation, bank 225 Capital asset pricing model (CAPM) 179–80, 252–5, 257–8 Capital market line (CML) 250–2 CAPM (capital asset pricing model) 179, 252–5, 257–8 CARA (constant absolute risk aversion) 233–4 Cartesian plane/space 39 Case studies Amex 144–6, 153–5 BHP Billiton Ltd 162–5, 168–70, 174–5, 177–8 credit risk 171–3 EM algorithm 203–6 PCA of European equity index 67–9 time series of asset price 218–20 Cauchy distribution 105 CBOE Gold index 162–3, 168–70, 174 Central limit theorem 120–1 Centre of location of distribution 78–9 Certainty equivalence 227–9 Characteristic equation 51–2 Chi-squared distribution 100–1, 123–4 Cholesky decomposition 37–8, 61–3, 70 Cholesky matrix 62–3, 70, 220–2 Circulant matrix 178 Classical probability 72–3 CML (capital market line) 250–2 Coefficients OLS estimators 155 regression 143–4, 151–2, 155, 168–9 risk aversion 231–4, 237 risk tolerance 233 Cokurtosis, CAPM 255 Complement, probability 73 Complete market 212 Compounding factor, returns 22–3 Concave function 13–14, 35 Conditional confidence interval 169 Conditional distribution 108–9 Conditional mean equation, OLS 148 Conditional probability 73 Conditional value at risk 105 Confidence interval 72, 118–24, 167–70 Conjugate gradient 193 Consistent OLS estimators 156–8 Constant absolute risk aversion (CARA) 233–4 Constant relative risk aversion (CRRA) 232–4 Constant term, regression 143–4 Constrained optimization 29–31 Constraint, minimum variance portfolio 245–6 Continuous compounding, return 22–3 Continuous distribution 114 Continuous function 5–6, 35 Continuous time 134–9 long-short portfolio 21 mean reverting process 136–7 Index notation 16–17 P&L 19 random walks 136–7 stochastic process 134–9 Convergence, iteration 188–9 Convex function 13–14, 35 Copula 109–10 Correlation 111–14 beta value 147–8 simulation 220–2 Correlation matrix 38, 55–61, 70 eigenvalues/vectors 52–4, 59–61 PCA 64–5, 67–8, 70 positive definiteness 58–9 Coskewness, CAPM 255 Counting process 139 Coupon 1 Covariance 80, 110–2 Covariance matrix 37–8, 55–61, 70 eigenvalues/vectors 59–61 OLS estimation 159–60 PCA 64, 66–7, 70 positive definiteness 58–9 Cox–Ross–Rubinstein parameterization 215 Crank–Nicolson finite difference scheme 210 Credit risk case study 171–3 Criterion for convergence, iteration 188 Critical region, hypothesis test 124–5 Critical value 118–20, 122–3, 129 Cross-sectional data 144 CRRA (constant relative risk aversion) 232–4 Cubic spline 197–200 Cumulative distribution function 75 Currency option 195–7 Decision rule, hypothesis test 125 Decomposition of matrix 61–4 Definite integral 15–16 Definite matrix 37, 46–7, 54, 58–9, 70 Degree of freedom, Student t distribution Delta–gamma approximation 2–3, 34 Delta–gamma–vega approximation 34 Delta hedging 208, 211 Density function 75–7 binomial distribution 86 bivariate distribution 108–9 joint 114–15 leptokurtic 82–3 lognormal distribution 93 MLE 130–4 97–8 normal distribution 90–2, 97, 115–17 Poisson distribution 88 stable distribution 105–6 Student t distribution 97–100 Dependent variable 143 Derivatives 1–2 calculation 12–13 first 2, 10–11 partial 27–8, 35 second 2, 11, 13 total 31 Determinant 41–3, 47 Deterministic variable 75 Diagonalizable matrix 43 Diagonal matrix 40, 56 Dicky–Fuller test 136 Differentiable function 5–6, 35 Differential equations partial 2, 208–10 stochastic 134, 136 Differentiation 10–15 concave/convex function 13–14 definition 10–11 monotonic function 13–14 rule 11–12 stationary point 14–15 stochastic differential term 22 Diffusion process, Brownian motion 22 Discontinuity 5 Discrete compounding, return 22–3 Discrete time 134–9 log return 19–20 notation 16–17 P&L 19 percentage return 19–20 random walk model 135 stationary/integrated process 134–6 stochastic process 134–9 Discretization of space 209–10 Discriminant 5 Distribution function 75–7 Diversifiable risk 181 DJIA (Dow Jones Industrial Average) index 137–8 Dot product 39 Dow Jones Industrial Average (DJIA) index 137–8 Dummy variable 175 Duration–convexity approximation 2–3, 34 Durbin–Watson autocorrelation test 176–7 281 282 Index Efficiency, OLS estimators 156–7 Efficient frontier 246–9, 251 Efficient market hypothesis 179 Eigenvalues/vectors 37–8, 48–54, 70 characteristic equation 51–2 correlation matrix 52, 59–60 covariance matrix 59–61 definiteness test 54 definition 50–1 linear portfolio 59–61 matrices/transformations 48–50 properties 52–3 Elliptical distribution 115 EM (expectation–maximization algorithm) 203–6 Empirical distribution 77, 217–18 Enhanced indexation 182–3 Epanechnikov kernel 107 Equality of two mean 126–7 Equality of two variance 126–7 Equations 3–9 CML 252 heat equation 208–9 partial differential 2, 208–10 quadratic 4–5 roots 187 simultaneous equations 44–5 Equity index returns 96–7 Equity price 172 Equity volatility 172–3 Error process 145, 148, 155 ESS (explained sum of squares) 149–50, 159–62 Estimation calibration 201 MLE 72, 130–4, 141, 202–3 OLS 143–4, 146–9, 153–63, 170–1, 176 ETL (expected tail loss) 104–5 European equity indices case study 67–9 European options 1–2 American option 215–16 binomial lattice 212–13 interpolation 195–6 pricing 212–13, 215–16 Euro swap rate (1-year) 172 Excel BHP Billiton Ltd case study 163–4 binomial distribution 213 chi-squared distribution 123–4 critical values 118–20, 122–3 Goal Seek 186, 188–9 histogram 77–8 matrix algebra 40, 43–6, 53–4, 59, 63–4, 68, 70 moments of a sample 82–3 multiple regression 163–4 normal probabilities 90–1 OLS estimation 153–5 percentiles 83–5 Poisson distribution 88 random numbers 89 simulation 217, 219 Solver 186, 190–1, 246 Student t distribution 100, 122–3 Expectation–maximization (EM) algorithm 203–6 Expected tail loss (ETL) 104–5 Expected utility 227–8 Expected value (expectation) 78–9 Explained sum of squares (ESS) 149–50, 159–62 Explanatory variables 143, 157, 170 Explicit function 185 Exponential distribution 87–9 Exponential function 1, 7–9, 34–5, 233–7 Exponential utility function 233–7 Extrapolation 186, 193–200, 223 Extreme value theory 101–3 Factorial function 8 Factorial notation 86 Factor model software 181 F distribution 100–1, 127 Feasible set 246 Finance calculus 1–36 essential algebra 37–70 numerical methods 185–223 portfolio mathematics 225–67 Financial market integration 180–1 Finite difference approximation 186, 206–10, 223 first/second order 206–7 the Greeks 207–8 partial differential equations 208–10 First difference operator, discrete time 17 First order autocorrelation 178 Forecasting 182, 254 Forward difference operator, returns 19, 22 Index Fréchet distribution 103 F test 127 FTSE 100 index 204–5, 242–4 Fully-funded portfolio 18 Functions 3–9, 26–31 absolute value 6 concave 13–14, 35 continuous 5–6, 35 convex 13–14, 35 differentiable 5–6, 35 distribution function 75–7, 114–15 explicit/implicit 185 exponential 1, 7–9, 34–5, 234–7 factorial 8 gamma 97–8 indicator 6 inverse 6–7, 35 Lagrangian 29–30 likelihood 72, 130–4 linear 4–5 logarithmic 1, 9, 34–5 monotonic 13–14, 35 non-linear 1–2 objective 29, 188 quadratic 4–5, 233–4 several variables 26–31 utility 232–4 Fundamental theorem of arbitrage 212 Future 1, 181–2, 194 Gamma function, Student t distribution 97–8 Gaussian copula 109–10 Gaussian kernel 107 Gauss–Markov theorem 157, 175, 184 Generalized extreme value (GEV) distribution 101–3 Generalized least squares (GLS) 178–9 Generalized Pareto distribution 101, 103–5 Generalized Sharpe ratio 262–3 General linear model, regression 161–2 Geometric Brownian motion 21–2 lognormal asset price distribution 213–14 SDE 134 stochastic process 141 time series of asset prices 218–20 GEV (generalized extreme value) distribution 101–3 Global asset management 225 Global minimum variance portfolio 244, 246–7 283 GLS (generalized least squares) 178–9 Goal Seek, Excel 186, 188–9 Gold index, CBOE 162–3, 168–70, 174 Goodness of fit 128, 149–50, 163–5, 167 Gradient vector 28 Graphs 3–9 The Greeks 207–8 Gumbel distribution 103 Heat equation 208 Hedging 2, 181–2 Hermite polynomials 200 Hessian matrix 28–30, 132, 192–3 Heteroscedasticity 175–9, 184 Higher moment CAPM model 255 Histogram 76–8 Historical data, VaR 106 Homoscedasticity 135 h-period log return 23–4 Hyperbola 5 Hypothesis tests 72, 124–5 CAPM 254–5 regression 151–2, 163–6 Student t distribution 100 Identity matrix 40–1 i.i.d.

How I Became a Quant: Insights From 25 of Wall Street's Elite
by Richard R. Lindsey and Barry Schachter
Published 30 Jun 2007

In a key lucky break, Goldman Sachs decided (at our prodding) to seed with partner capital a very aggressive market-neutral hedge fund utilizing our new investment process.14 Although we had very strong results in general across many products (both long only and absolute return), over the next few years our results for this hedge fund were off the charts. These results were not just great, but much better than our own backtests, a key sign you’re getting at least somewhat lucky as an iron-clad rule is to expect results worse than your backtest. Don’t get me wrong, I think we created some great models, but getting a lucky draw on top of a great model is a pretty wonderful thing to happen early in your career. (As they say in the novel Dune, “Beginnings are delicate times.”)15 A few years down the road, we were managing $7 billion, about $6 billion in long-only assets, and close to a billion in hedge fund assets, all with strong-to-stellar results.

…

The focus of the fund was to create automated trading strategies and apply them to global futures markets, including commodities, equities, and fixed income. As long as it was a valid futures market, we traded it, regardless if the prices represented Eurodollar contracts or Red Azuki Beans. I spent a lot of time writing very complex code to create and backtest different types of trading strategies using daily futures data back to the 1940s. Oodles of data, challenging analyses, and lots of programming— this is exactly what I had been doing in physics for a dozen years, and I was groovin’. But alas, I quickly came to realize that finance is not rocket science.

…

As a two-person operation, my first task was simple—recreate, from scratch, everything that the previous 30-person fund had done, but in a way that could be wholly automated and required no additional staff. Over that next year I coded day and night, and even purchased a $20,000 Sun SparcStation laptop (that’s right, a laptop) so I could code during my two-hour-per-day train commute. I was in heaven. I created my own futures backtesting language, a byte-code compiler, and an automated web-based trading system. With these tools I developed many new styles of trend following that had never been done before at the previous fund. Each night the system would upload the latest closing prices for each futures market, rerun my simulation routines, generate signals, and auto-fax trades to our London brokers for execution the next morning.

pages: 467 words: 154,960

Trend Following: How Great Traders Make Millions in Up or Down Markets
by Michael W. Covel
Published 19 Mar 2007

The activities of a risk manager are to measure risk and to increase and decrease risk by buying and selling stock. In general, good risk management combines several elements: 1. Clarifying trading and risk management systems until they can translate to computer code. 2. Inclusion of diversification and instrument selection into the back-testing process. 3. Back-testing and stress-testing to determine trading parameter sensitivity and optimal values. 4. Clear agreement of all parties on expectation of volatility and return. 5. Maintenance of supportive relationships between investors and managers. 6. Above all, stick to the system. 7. See #6, above.

…

Edward de Bono This page intentionally left blank Trading System Example from Mechanica F “Part of back-testing is to determine position sizing and risk management strategies that fit within your drawdown tolerance envelope.” —Ed Seykota1 In this appendix, Bob Spear shows how a trader might construct a simple, mechanical trend following system on Trading Recipes Portfolio Engineering Software. His newest software, surpassing Trading Recipes, is called Mechanica (www.mechanicasoftware. com). For this example we start with a broad look at the system’s trading ideas, which echo many of the ideas discussed in this book. We construct a hypothetical portfolio and run a backtest up to a certain point in time.

…

He was generous with his ideas, making a point to share what he knew; it delighted him to get others to try systems. He inspired a great many people and spawned a whole generation of traders, providing courage and a road map. Ed Seykota97 We started our database using punch cards in 1968, and we collected commodity price data back to July 1959. We back-tested the 5 and 20 and the weekly rules for Dick. I think the weekly method was the best thing that anyone had ever done. Of all Dick’s contributions, the weekly rules helped identify the trend and helped you act on it. Dick is one of those people who today likes to beat the computer—only he did it by hand.

pages: 314 words: 122,534

The Missing Billionaires: A Guide to Better Financial Decisions
by Victor Haghani and James White
Published 27 Aug 2023

This is the same index that we used (although in the opposite direction) in the backtest we discussed earlier that looked at the performance of being long the stock market protected with these one month, 2% out‐of‐the‐money put options. The comparable strategy without options would be to keep 50% (roughly the time average beta of the put‐selling strategy in our sample) of wealth in the S&P 500 and the rest in T‐bills, since this gives approximately the same average expected exposure as investors get from selling puts on 100% of their wealth. Over the same period as our other backtest, we found that the returns of the option‐selling strategy are about the same as for the simpler strategy of keeping 50% in the stock market.

…

Finally, while it did not occur during the 30‐year sample, the PUTY trade explicitly exposes investors to the risk of losing all their money if the stock market goes down 100%, whereas such a decline in the stock market wouldn't be as catastrophic for investors who allocated 50% to stocks.k This is all to say that the Risk‐adjusted Return of the 50% in the S&P 500 strategy was significantly higher than that of the put‐selling strategy. Of course, 30 years is a short sample, and we do not suggest that an investor make a decision solely on the basis of this limited backtest. But if your prior belief is that put‐selling doesn't make sense for individual investors, this backtest should give you moderately greater confidence in it. It may be that a small dose of short volatility strategies would improve investor welfare, but not significantly, especially after accounting for taxes and the transaction, monitoring, and management costs involved in the put‐selling strategy.

…

But such as it is, history does support the hypothesis that options on the stock market are generally expensive, which suggests that out‐of‐the‐money puts are expensive as well. For example, over the 10 years to the end of 2021, the average value of the VIX index of stock option implied volatility has been 3.6% higher than the relevant realized volatility.9 Our simple backtest also agrees with the majority view among academics that, over a long‐term cycle, options are priced somewhat rich to fair value.10 There has been much heated debate about whether it is possible to buy stock market insurance at a fair or even cheap price. There are a number of purveyors of “tail hedging” investment strategies who contend that they do provide Portfolio Insurance at a discount to its fair value.11 But other experts argue it's more likely that stock market insurance is priced at a premium to fair value.12 Also, for investors who care more about the risk of long‐term underperformance than of short‐term gaps moves, evidence suggests that trend‐following strategies have been able to provide effective stock portfolio protection at much lower cost.h A limited backtest over the 30 years from 1991 to the end of 2021 comparing a strategy of being 100% invested in the US stock market plus owning 2% out‐of‐the‐money one‐month puts versus a simpler strategy of keeping 50% in the stock market and 50% in T‐bills found the strategy with puts generates a lower return over the period, with higher risk and a lower Sharpe ratio.i There are some periods over which the strategy with puts would have generated a smaller drawdown, but the maximum drawdown for the two strategies was almost the same.

pages: 156 words: 15,746

Personal Finance with Python
by Max Humber

prices = historical_prices.loc['2016-01-04'] prices AMZN 636.99 CSCO 26.41 GE 30.71 Name: 2016-01-04 00:00:00, dtype: float64 Portfolio Because I know you just want to see the full thing in action, here it is: portfolio = instantiate_portfolio(targets, 100000.00) prices = historical_prices.loc['2017-01-01'] update_prices(portfolio, prices) order = get_order(portfolio) simulate_process_order(portfolio, order) portfolio.market_value.sum() 100000.0 This will be our starting portfolio: print(portfolio) date price target allocation shares market_value AMZN 2017-01-01 749.87 0.4 0.397431 53 39743.1 CSCO 2017-01-01 30.22 0.3 0.299782 992 29978.2 GE 2017-01-01 31.6 0.3 0.299884 949 29988.4 CASH 2017-01-01 1 0 0.0029025 290.25 290.25 Rebalance To test our rebalancing logic, we’ll back-test across 2017 and execute orders on a quarterly-end frequency by using the Q offset alias from pandas. dates = pd.date_range('2017-01-01', '2017-12-31', freq="Q").tolist() for d in dates: prices = historical_prices.loc[d] update_prices(portfolio, prices) order = get_order(portfolio) print(f'{d}:\n{order}') simulate_process_order(portfolio, order) portfolio.market_value.sum() 2017-03-31 00:00:00: AMZN -4 CSCO -24 GE 149 dtype: object 2017-06-30 00:00:00: AMZN -5 CSCO 63 GE 97 dtype: object 2017-09-30 00:00:00: AMZN 0 CSCO -83 GE 124 dtype: object 2017-12-31 00:00:00: AMZN -7 CSCO -79 GE 589 dtype: object 111030.14 After four rebalancing moves, we can verify that our portfolio will follow and maintain target allocations quite closely.

…

print(portfolio) date price target allocation shares market_value AMZN 2017-12-31 1169.47 0.4 0.389718 37 43270.4 CSCO 2017-12-31 38.3 0.3 0.299763 869 33282.7 GE 2017-12-31 17.45 0.3 0.29987 1908 33294.6 CASH 2017-12-31 1 0 0.0106498 1182.45 1182.45 Conclusion In this chapter, you learned how to build a portfolio in pandas, update values in a DataFrame, generate buy and sell orders that aim to hold target allocations in balance, retrieve stock quotes from Alpha Vantage, and simulate back-testing. If you want to actually put these pieces to work, you will need to set up an account with an online brokerage and manually exercise buy and sell orders on its platform. The good news is that if you think that rebalancing is an appropriate investment strategy for you, you don’t actually have to do it that often.

pages: 502 words: 107,657

Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die
by Eric Siegel
Published 19 Feb 2013

With secrecy reigning supreme, the protocol for this type of audit dictated that John receive only the numerical results, along with a few adjectives that described its design: new, unique, powerful! With meager evidence, John sought to prove a crime he couldn’t even be sure had been committed. Before each launch, organizations establish confidence in PA by “predicting the past” (aka backtesting). The predictive model must prove itself on historical data before its deployment. Conducting a kind of simulated prediction, the model evaluates across data from last week, last month, or last year. Feeding on input that could only have been known at a given time, the model spits out its prediction, which then matches against what we now already know took place thereafter.

…

On a hunch, he hand-crafted a method with the same type of bug, and showed that its predictions closely matched those of the trading system. A predictive model will sink faster than the Titanic if you don’t seal all its “time leaks” before launch. But this kind of “leak from the future” is common, if mundane. Although core to the very integrity of prediction, it’s an easy mistake to make, given that each model is backtested over historical data for which prediction is not, strictly speaking, possible. The relative future is always readily available in the testing data, easy to inadvertently incorporate into the very model trying to predict it. Such temporal leaks achieve status as a commonly known gotcha among PA practitioners.

…

And in the commercial world, John’s pregrad positions had dropped him right into the thick of machine learning systems that steer for aerospace flight and that detect cooling pipe cracks in nuclear reactors, not to mention projects for Delta Financial looking over the shoulders of other black box quants. And now John’s latest creation absolutely itched to be deployed. Backtesting against historical data, all indications whispered confident promises for what this thing could do once set in motion. As John puts it, “A slight pattern emerged from the overwhelming noise; we had stumbled across a persistent pricing inefficiency in a corner of the market, a small edge over the average investor, which appeared repeatable.”

pages: 1,088 words: 228,743

Expected Returns: An Investor's Guide to Harvesting Market Rewards
by Antti Ilmanen
Published 4 Apr 2011

This outcome may be a coincidence but it echoes the finding that January equity market performance has some predictive ability for rest-of-the-year returns (further details in Chapter 25). Incorporating these two seasonal biases would easily improve backtested FX carry strategy performance—for example, doubling position sizes for January and halving sizes for the rest of the year if the January return had been negative would have boosted the Sharpe ratio since 1983 from 0.6 to 0.8. Any such backtest improvements are subject to data-mining bias, so some skepticism is warranted. Conditioners (regime indicators) As we have seen, ex ante opportunities and seasonal effects have some ability to predict carry returns.

…

• For specific funds and strategies, the historical performance data that investors get to see are often upward biased. This bias is due to the voluntary nature of performance reporting and survivorship bias (so that poor performers are left out of databases or are not marketed by the fund manager). A similar caveat applies to simulated “paper” portfolios because backtests may be overfitted and trading costs ignored or understated. These concerns notwithstanding, this book presents extensive evidence of long-run realized returns, when possible covering 50-to-100-year histories. Several main findings are familiar to most readers:• Stock markets have outperformed fixed income markets during the past century in all countries studied.

…

For active asset managers with voluntary reporting, published returns are almost certainly upward biased. Section 11.4 reviews a host of selection biases such as survivorship bias and backfill bias in the context of hedge fund return databases, but similar caveats apply to the reported performance of other managers. Backtested results of active strategies also suffer from overfitting and data-mining biases, which also overstate published returns. Whenever we observe exceptionally attractive historical returns, it is healthy to adopt a skeptical approach. The financial industry has limited incentives to emphasize this needed skepticism beyond printing required disclaimers, while our innate tendencies for extrapolation and optimism make most of us too easy prey for the upbeat marketing of past performance. 4.7 NOTES [1] The distinction between realized (ex post) and expected (ex ante) returns should be crystal clear.

pages: 111 words: 1

Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets
by Nassim Nicholas Taleb
Published 1 Jan 2001

This simple remark, possibly blurted out in a state of intoxication or extreme enthusiasm, caused Dos Passos to become required reading in European intellectual circles, as Sartre’s remark was mistaken for a consensus estimate of the quality of Dos Passos rather than what it was, the best remark. (In spite of such interest in his work, Dos Passos has reverted to obscurity.) The Backtester A programmer helped me build a backtester. It is a software program connected to a database of historical prices, which allows me to check the hypothetical past performance of any trading rule of average complexity. I can just apply a mechanical trading rule, like buy NASDAQ stocks if they close more than 1.83% above their average of the previous week, and immediately get an idea of its past performance.

…

A Mischievous Child Replaces the Black Balls Seven THE PROBLEM OF INDUCTION FROM BACON TO HUME Cygnus Atratus Niederhoffer SIR KARL’S PROMOTING AGENT Location, Location Popper’s Answer Open Society Nobody Is Perfect Induction and Memory Pascal’s Wager THANK YOU, SOLON PART II: MONKEYS ON TYPEWRITERS • Survivorship and Other Biases IT DEPENDS ON THE NUMBER OF MONKEYS VICIOUS REAL LIFE THIS SECTION Eight TOO MANY MILLIONAIRES NEXT DOOR HOW TO STOP THE STING OF FAILURE Somewhat Happy Too Much Work You’re a Failure DOUBLE SURVIVORSHIP BIASES More Experts Visibility Winners It’s a Bull Market A GURU’S OPINION Nine IT IS EASIER TO BUY AND SELL THAN FRY AN EGG FOOLED BY NUMBERS Placebo Investors Nobody Has to Be Competent Regression to the Mean Ergodicity LIFE IS COINCIDENTAL The Mysterious Letter An Interrupted Tennis Game Reverse Survivors The Birthday Paradox It’s a Small World! Data Mining, Statistics, and Charlatanism The Best Book I Have Ever Read! The Backtester A More Unsettling Extension The Earnings Season: Fooled by the Results COMPARATIVE LUCK Cancer Cures Professor Pearson Goes to Monte Carlo (Literally): Randomness Does Not Look Random! The Dog That Did Not Bark: On Biases in Scientific Knowledge I HAVE NO CONCLUSION Ten LOSER TAKES ALL—ON THE NONLINEARITIES OF LIFE THE SANDPILE EFFECT Enter Randomness Learning to Type MATHEMATICS INSIDE AND OUTSIDE THE REAL WORLD The Science of Networks Our Brain Buridan’s Donkey or the Good Side of Randomness WHEN IT RAINS, IT POURS Eleven RANDOMNESS AND OUR MIND: WE ARE PROBABILITY BLIND PARIS OR THE BAHAMAS?

…

If enough trading rules are considered over time, some rules are bound by pure luck, even in a very large sample, to produce superior performance even if they do not genuinely possess predictive power over asset returns. Of course, inference based solely on the subset of surviving trading rules may be misleading in this context since it does not account for the full set of initial trading rules, most of which are unlikely to have underperformed. I have to decry some excesses in backtesting that I have closely witnessed in my private career. There is an excellent product designed just for that, called Omega TradeStation, that is currently on the market, in use by tens of thousands of traders. It even offers its own computer language. Beset with insomnia, the computderized day traders become night testers plowing the data for some of its properties.

Trading Risk: Enhanced Profitability Through Risk Control
by Kenneth L. Grant
Published 1 Sep 2004

Specifically, all models will yield results that have a probabilistic component, each producing a number that serves as a threshold that the portfolio volatility should violate with a specified level of probability. As such, the accuracy of the model can be back-tested by checking how many P/L observations actually exceed the VaR figure over a given period of time and then tying the resulting percentage back to the confidence interval. If the two figures are consistent, the model can be said to be in good order; if not, then the model’s assumptions and inputs must be examined in order to determine the source of the inconsistency (more about this later). This type of backtest is one of the simplest mathematical exercises we will cover in this book, and I urge you to periodically check to see if your actual P/L volatility is consistent with the model and parameters you are using.

…

This type of backtest is one of the simplest mathematical exercises we will cover in this book, and I urge you to periodically check to see if your actual P/L volatility is consistent with the model and parameters you are using. Some purists will argue that comparing actual portfolio volatility to VaR results is not an effective backtest because while the VaR calculation is based on the current portfolio as measured against historical price action, the test measures the accuracy of this calculation against future price fluctuation. 99 The Risk Components of an Individual Portfolio However, I would ignore this distinction for the simple reason that if your actual P/L volatility differs substantially from the results implied in the VaR calculation, it is failing to tell you anything—irrespective of the fact that it may be faithfully adhering to its core assumption base.

…

Although the same The Risk Components of an Individual Portfolio 103 dollars are committed to the market from each portfolio, I would argue that the high-flying portfolio is more fully invested and has a higher degree of market participation, largely due to the fact that it will experience more acute financial impacts to any level of price action in the markets than will the dividend portfolio. VaR has no purpose other than to measure these concepts, and I particularly endorse your taking the steps necessary to understand the time series dynamics of the VaR calculation, starting with a routine periodic measurement of its predictive accuracy (i.e., the back-test). Study its ability to adhere to the confidence interval assumptions you have specified (as described in the previous Testing VaR Accuracy section). Once you have achieved a sufficient level of accuracy in this regard, you can then use the VaR statistic as yet another means of ensuring the efficient use of your capital.

Hedgehogging
by Barton Biggs
Published 3 Jan 2005

Ned Davis says, “Go with the flow until it reaches an extreme and begins to reverse; it is at ccc_biggs_ch10_133-148.qxd 11/29/05 7:02 AM Page 145 The Battle for Investment Survival 145 that point where it often pays to be contrary.” Merrill Lynch did a fascinating study that back-tested the four main sentiment indicators and found that they had virtually no predictive powers at market peaks, although they were decent at bottoms. Morgan Stanley did a similar analysis of the VIX Index and came to a similar conclusion.This index, which measures complacency and which everyone tracks on a daily basis, currently has fallen to a very low level. That suggests a high degree of complacency and is, therefore, a bearish sign. However, when Morgan Stanley back-tested its predictive ability, it found VIX virtually useless in signaling tops.

…

ccc_biggs_ch02_9-20.qxd 11/29/05 6:57 AM Page 11 Golden Boys, but They Still Bleed Red 11 Sorry, but I think market neutral is a tough racket, particularly quantitative market neutral.There are too many people doing the same thing. In the 1980s, Morgan Stanley had a series of market-neutral funds that were run off different fundamental and quantitative models.All had boy geniuses at the controls who had made big money on the trading desk or had built a model that worked great on paper when they were dummy back-tested. In live action, none ever produced with real money. One allegedly used a computer to take stock selection to the third derivative and fifth dimension in color. There was color all right. The color of the P&L was a deep shade of red. Anyway, this attractive trio from Goldman opened on January 1, 2001, with $100 million and lost only 10% over the next two years, which wasn’t bad considering the bear market.Then in 2003, they were too cautious and bearish, and the fund went up only 5%.

…

My only consolation is that this disdain, although a bad omen in terms of money coming in, is probably a good sign for our initial performance. This evening, at the dinner for us organized by Prime Brokerage, we made our usual presentation and then took questions. As always, we said we hoped to have long-term returns in excess of the S&P 500, but that our back-testing indicated volatility approximately comparable to that of the S&P 500. In other words, if our long-term rate of return is going to be about 15% per annum, our annual returns should fall in a range of +30% to zero—in other words, 15%, plus or minus 15%. A woman from a big fund of funds stood up and said she thought our volatility estimate of 15% was about right, but it made us unsuitable for many investors.

pages: 512 words: 162,977

New Market Wizards: Conversations With America's Top Traders
by Jack D. Schwager
Published 28 Jan 1994

I would like to say that the procedure was immensely successful, but the truth is that the person didn’t experience any overnight transformation. It took many years before I realized why hypnosis was very effective with some traders but not others. What is the reason? Some traders have a valid methodology that they have adequately backtested and that their conscious mind is happy with. These are the traders who can usually be helped through hypnosis. The only thing hypnosis can do is to inform the subconscious mind that the person now has a valid methodology that the conscious mind has already accepted. But you must first be at that point.

…

Some people lose because they feel they don’t deserve to win, but more people lose because they never perform the basic tasks necessary to become a winning trader. What are those tasks? 1. 2. 3. 4. 5. Develop a competent analytical methodology. Extract a reasonable trading plan from this methodology. Formulate rules for this plan that incorporate money management techniques. Back-test the plan over a sufficiently long period. Exercise self-management so that you adhere to the plan. The best plan in the world cannot work if you don’t act on it. Typically, how do you work with someone who comes to you for help in improving his or her trading? The first thing I do is go though a series of about thirty questions that have only one purpose: finding out if the person has a methodology.

…

However, the losses from your previous Methodology A are so ingrained in your subconscious that whenever you contemplate making a trade, the adrenaline starts to flow, and the fear of executing a trade arises. Some traders are literally immobilized by this fear at the moment when they need to act. This is the “freeze” that I encountered when I returned to trading years after my first painful experience. If you have truly back-tested a methodology and are employing an effective trading plan, your conscious mind is already aware of its validity. It’s your subconscious mind that prevents you from taking correct action in the market. The problem will persist until you convince the subconscious in a very direct manner that the new methodology is valid and that it has to forget about the old methodology.

pages: 272 words: 19,172

Hedge Fund Market Wizards
by Jack D. Schwager
Published 24 Apr 2012

When I liquidated the trade, I would look at what actually happened and compare it with my reasoning and expectations when I put on the trade. Learning solely from actual experience, however, is inadequate because it takes too much time to get a representative sample to determine whether a decision rule works. I discovered that I could backtest the criteria that I wrote down to get a good perspective of how they would have performed and to refine them. The next step was to define decision rules based on the criteria. I required the decision rules to be logically based and was careful to avoid data mining. That’s how the Bridgewater system began and developed in the early years.

…

I started thinking, This place is going to empty out sometime tonight. I decided to get all my data prepared so that I could simultaneously use many of the lab’s computers that night. I was very excited about the idea. People started to leave, and then I had two computers, then four, and eventually I was jumping between 20 computers running my backtests. Were you testing your system on one market on each computer? That is exactly what I was doing. I was so excited about the results I was getting that I worked all night and continued through the next day. It was going so well that I pulled a second all-nighter. I worked for nearly 40 hours straight, keeping myself awake with the caffeine from drinking a Pepsi every hour.

…

By 1994, I was trading about 20 markets, and I was no longer using market-specific models. Those changes made a big difference. When you were only trading two or three markets, how did you decide which markets to trade? That was part of the problem. I was cherry-picking the markets that looked best in backtesting. It sounds like you were still making some rookie curve-fitting mistakes at that time. Absolutely. I was still making some very bad data mining errors in those initial years. Was the system you were using at Blue Ridge after you switched to using the same models on all markets an early version of what you ended up doing at QIM?

pages: 407 words: 104,622

The Man Who Solved the Market: How Jim Simons Launched the Quant Revolution
by Gregory Zuckerman
Published 5 Nov 2019

A golden age for traditional investing had dawned as George Soros, Peter Lynch, Bill Gross, and others divined the direction of investments, financial markets, and global economies, producing enormous profits with intelligence, intuition, and old-fashioned economic and corporate research. Unlike his rivals, Simons didn’t have a clue how to estimate cash flows, identify new products, or forecast interest rates. He was digging through reams of price information. There wasn’t even a proper name for this kind of trading, which involved data cleansing, signals, and backtesting, terms most Wall Street pros were wholly unfamiliar with. Few used email in 1990, the internet browser hadn’t been invented, and algorithms were best known, if at all, as the step-by-step procedures that had enabled Alan Turing’s machine to break coded Nazi messages during World War II. The idea that these formulas might guide, or even help govern, the day-to-day lives of hundreds of millions of individuals, or that a couple of former math professors might employ computers to trounce seasoned and celebrated investors, seemed far-fetched if not outright ludicrous.

…

The roots of Simons’s investing style reached as far back as Babylonian times, when early traders recorded the prices of barley, dates, and other crops on clay tablets, hoping to forecast future moves. In the middle of the sixteenth century, a trader in Nuremberg, Germany, named Christopher Kurz won acclaim for his supposed ability to forecast twenty-day prices of cinnamon, pepper, and other spices. Like much of society at the time, Kurz relied on astrological signs, but he also tried to back-test his signals, deducing certain credible principles along the way, such as the fact that prices often move in long-persisting trends. An eighteenth-century Japanese rice merchant and speculator named Munehisa Homma, known as the “god of the markets,” invented a charting method to visualize the open, high, low, and closing price levels for the country’s rice exchanges over a period of time.

…

See Statistical arbitrage Archimedes (yacht), 267, 320 Armstrong, Neil, 170 Artin, Emil, 69 Asness, Clifford, 256–57 Association for Computing Machinery (ACM), 37 astrology, 121–22 autism, xviii, 268, 287, 323–24 Automated Proprietary Trading (APT), 131–32, 133 AWK, 233–34 Ax, Frances, 98 Ax, James, xi, 37, 68–69, 324 at Axcom Limited, 78–83 backgammon, 69, 76–77 background of, 68–69 at Berkeley, 68–69 Berlekamp and, 95–102 conspiracy theories of, 77–78, 99 at Cornell, 69, 70–71 death of, 103 focus on mathematics, 69–70 at Monemetrics, 51–52, 72–73 personality of, 68, 70, 71–72, 98–99 Simons and, 34, 68–69, 99–103, 107 at Stony Brook, 34, 71–72 trading models, 73, 74–75, 77–78, 81–86, 95–101, 107 Axcom Limited, 78–83 disbanding of, 118 trading models, 95–101, 107–18 Ax-Kochen theorem, 69, 70, 103 Bachelier, Louis, 128 backgammon, 69, 76–77 backtesting, 3 Bacon, Louis, 140 Baker House, 15–16 Baltimore City Fire and Police Employees’ Retirement System, 299–300 Bamberger, Gerry, 129–30 BankAmerica Corporation, 212 Bannon, Steve, 279, 280, 280n break with Mercers, 304 at Breitbart, 278–79, 299–300, 301–2 midterm elections of 2018, 304 presidential election of 2016, xviii, 281–82, 284–85, 288–90, 293, 294–95 Barclays Bank, 225, 259 bars, 143–44 Barton, Elizabeth, 272 basket options, 225–27 Baum, Julia Lieberman, 46, 48, 50, 62–63, 65 Baum, Leonard “Lenny,” xi, 45–46, 63–66 background of, 46 currency trading, 28–29, 49–53, 54–60, 62–64, 73 death of, 66 at Harvard, 46 at IDA, 25, 28–29, 46–49, 81 at Monemetrics, 45, 49–60, 63–65 move to Bermuda, 64–65 rift with Simons, 63–65 trading debacle of 1984, 65, 66 Baum, Morris, 46 Baum, Stefi, 48, 62, 63 Baum–Welch algorithm, 47–48, 174, 179 Bayes, Thomas, 174 Bayesian probability, 148, 174 Beane, Billy, 308 Beat the Dealer (Thorp), 127, 163 Beautiful Mind, A (Nasar), 90 behavioral economics, 152, 153 Bell Laboratories, 91–92 Belopolsky, Alexander, 233, 238, 241, 242, 252–54 Bent, Bruce, 173 Berkeley Quantitative, 118 Berkshire Hathaway, 265, 309, 333 Berlekamp, Elwyn, xi at Axcom, 94–97, 102–3, 105–18 background of, 87–90 at Bell Labs, 91–92 at Berkeley, 92–93, 95, 115, 118, 272 at Berkeley Quantitative, 118 death of, 118 at IDA, 93–94 Kelly formula and, 91–92, 96, 127 at MIT, 89–91 Simons and, 2–3, 4, 93–95, 109–10, 113–14, 116–18, 124 trading models and strategies, 2–3, 4, 95–98, 106–18, 317 Berlekamp, Jennifer Wilson, 92 Berlekamp, Waldo, 87–88 Berlin Wall, 164 Bermuda, 64–65, 254 Bernard L.

pages: 232 words: 70,835

A Wealth of Common Sense: Why Simplicity Trumps Complexity in Any Investment Plan
by Ben Carlson
Published 14 May 2015

Back to the investment conference: I found myself nodding in agreement with this fund manager as he surgically laid out the reasoning behind the potential shift to make better investment strategies available at a lower cost to more and more investors—increased competition, availability of information, a dearth of academic studies on back-tested strategies, and the fact that most professional portfolio managers came from similar schools of thought. This was making it harder and harder for portfolio managers to justify their claims of superior investment processes at a much higher cost to the individual investor. The line of thinking was that these newer products wouldn't offer the possibility for enormous outsized gains, but at a reduced cost to the investor, would give similar returns on a net basis after costs, the only thing that really matters in the end.

…

The majority of the time you should be like the pretty girl at the bar constantly turning down advances from the guys trying to hit on her. Be very selective about the phone numbers you'll take. In fact, a search for perfection can only lead to problems. You could go through all of the asset allocation studies, Monte Carlo simulations, and optimized portfolio back-tests of performance results you can find but all they will tell you is how certain portfolios have performed in the past. While these tools can be useful as a way to gauge possible risk factors, assuming future cycles will play out exactly as they have in the past can lead to overreactions when things don't go as planned.

pages: 825 words: 228,141

MONEY Master the Game: 7 Simple Steps to Financial Freedom
by Tony Robbins
Published 18 Nov 2014

Which is why I wasn’t the least bit surprised to learn later that he and his wife, Barbara, have signed the Giving Pledge—a commitment by the world’s wealthiest individuals, from Bill Gates to Warren Buffett, to give away the majority of their wealth through philanthropy. DO I HAVE YOUR ATTENTION NOW? When my own investment team showed me the back-tested performance numbers of this All Seasons portfolio, I was astonished. I will never forget it. I was sitting with my wife at dinner and received a text message from my personal advisor, Ajay Gupta, that read, “Did you see the email with the back-tested numbers on the portfolio that Ray Dalio shared with you? Unbelievable!” Ajay normally doesn’t text me at night, so I knew he couldn’t wait to share. As soon as our dinner date was over I grabbed my phone and opened the email . . .

…

By using real fund data as opposed to theoretical data from a constructed index, all the returns listed in this chapter are fully inclusive of annual fund fees and any tracking error present in the underlying funds. This has the benefit of showing you realistic historic returns for the All Seasons portfolio (as opposed to theoretical returns that are sometimes used in back-testing). This insures that the investment holdings and numbers used in back-testing this portfolio were and are accessible to the everyday man on the street and not only available to multibillion-dollar Wall Street institutions. Where they were unable to use actual index fund data because the funds didn’t exist at that time, they used broadly diversified index data for each asset class and adjusted the returns for fund fees.

…

If you want to take immediate action to minimize fees and have an advisor assist you in allocating your 401(k) fund choices, you can use the service at Stronghold (www.strongholdfinancial.com), which, with the click of a button will automatically “peer into” your 401(k) and provide a complimentary asset allocation. In addition, many people think there aren’t many alternatives to TDFs, but in section 5, you’ll learn a specific asset allocation from hedge fund guru Ray Dalio that has produced extraordinary returns with minimal downside. When a team of analysts back-tested the portfolio, the worst loss was just 3.93% in the last 75 years. In contrast, according to MarketWatch, “the most conservative target-date retirement funds—those designed to produce income—fell on average 17% in 2008, and the riskiest target-date retirement funds—designed for those retiring in 2055—fell on average a whopping 39.8%, according to a recent report from Ibbotson Associates.”

pages: 472 words: 117,093

Machine, Platform, Crowd: Harnessing Our Digital Future
by Andrew McAfee and Erik Brynjolfsson
Published 26 Jun 2017

The company faced the daunting task of building a technology platform for quants comparable to the ones within the industry’s top companies. Such a platform had to be able to let investors upload their algorithms, then quickly test them under different market conditions—booms and recessions, periods of high and low interest rates, and so on. One way to do this is to “back-test” the algorithms with historical data. Fawcett and his colleagues worked to build a backtester as robust as those available within large institutional investors. The startup also had to let investors accurately assess the market impact of their trades—the fact that if they bought or sold a large amount of an asset, that action would itself change the asset’s price.

…

We find Quantopian fascinating because it illustrates all three of the technology trends that are reshaping the business world. It’s bringing together minds and machines in fresh ways to rethink how investment decisions are made, and substituting data and code for human experience, judgment, and intuition. It’s also building a platform for quantitative investment rather than introducing a specific product (such as a backtester). This platform is open and noncredentialist, aims to take advantage of network effects (the more good investment algorithms it holds, the more capital it will attract; the more capital it holds, the more algo traders it will attract), and to provide a smooth interface and experience to its traders.

pages: 517 words: 139,477

Stocks for the Long Run 5/E: the Definitive Guide to Financial Market Returns & Long-Term Investment Strategies
by Jeremy Siegel
Published 7 Jan 2014

Index Options Buying Index Options Selling Index Options The Importance of Indexed Products Chapter 19 Market Volatility The Stock Market Crash of October 1987 The Causes of the October 1987 Crash Exchange Rate Policies The Futures Market Circuit Breakers Flash Crash—May 6, 2010 The Nature of Market Volatility Historical Trends of Stock Volatility The Volatility Index The Distribution of Large Daily Changes The Economics of Market Volatility The Significance of Market Volatility Chapter 20 Technical Analysis and Investing with the Trend The Nature of Technical Analysis Charles Dow, Technical Analyst The Randomness of Stock Prices Simulations of Random Stock Prices Trending Markets and Price Reversals Moving Averages Testing the Dow Jones Moving-Average Strategy Back-Testing the 200-Day Moving Average Avoiding Major Bear Markets Distribution of Gains and Losses Momentum Investing Conclusion Chapter 21 Calendar Anomalies Seasonal Anomalies The January Effect Causes of the January Effect The January Effect Weakened in Recent Years Large Stock Monthly Returns The September Effect Other Seasonal Returns Day-of-the-Week Effects What’s an Investor to Do?

…

Using historical data, it is impossible to determine times when the market average penetrated the 200-day moving average during the day. By specifying that the average must close above or below the average of the two hundred preceding closes, I present a theory that could have been implemented in practice through the whole time period.12 Back-Testing the 200-Day Moving Average Figure 20-2 displays the daily and 200-day moving averages of the Dow Jones Industrial Average during two select periods: from 1924 to 1936 and 2001 to 2012. The time periods when investors are out of the stock market (and in short-term bonds) are shaded; otherwise, investors are fully invested in stocks.

…

See also Calendar anomalies, 336 The Monetary History of the United States , 33–34, 212 Monetary policy capital gains taxes in, 226 conclusions about, 226–227 corporate taxes, 224–225 Federal Reserve and, 213–214, 217 gold standard and, 213–217 inflation in, 220–226 interest rates in, 223, 225–226 introduction to, 209–210 market outbreaks and, 244 money in circulation and, 210–213 postdevaluation, 215–216 postgold, 216–217 prices in, 210–213 stocks and inflation in, 220–226 supply-side effects in, 223–224 Money creation, 217 Money in circulation, 210–213 Money managers, 363–364 Money Market Investor Funding Facility, 33 Monsanto, 205 Moody’s, 25–26 Moore, Philip, 371 Moore, Randell, 236 Morgan, J.P., 125 Morgan Stanley Capital International (MSCI), 13, 42, 330–332, 371–372 Morris, David, 371 Mortgage-backed securities, 25–31 Moving averages 200-day, 318–320 back-testing, 318 bear market avoidance in, 320–321 Dow Jones strategy for, 317–318 gains/losses distributions in, 321–322 introduction to, 316 testing, 317–318 MSCI (Morgan Stanley Capital International), 13, 42, 330–332, 371–372 Mutual funds from 1995–2012, 272 equity, 358–363 ETFs vs., 273 index, 282–284 Mutual Shares Z Fund, 362 Myopic loss aversion, 350–352 Nabisco Group Holdings, 129–130 NAR (National Association of Realtors), 29 Nasdaq in 1999, 16–17 global investing and, 200–201 introduction to, 105 New York Stock Exchange and, 111–113 S&P 500 Index and, 120 National Association of Realtors (NAR), 29 National Association of Securities Dealers Automated Quotations (Nasdaq).

pages: 516 words: 157,437

Principles: Life and Work
by Ray Dalio
Published 18 Sep 2017

We certainly wouldn’t have done as well in the markets as we have. As you will see later, I am now developing similar systems to help us make management decisions. I believe one of the most valuable things you can do to improve your decision making is to think through your principles for making decisions, write them out in both words and computer algorithms, back-test them if possible, and use them on a real-time basis to run in parallel with your brain’s decision making. But I’m getting ahead of myself. Let’s go back to 1983. RESURRECTING BRIDGEWATER By late 1983, Bridgewater had six employees. Up until then, I hadn’t done any marketing; the business we got came from word of mouth and from people reading my daily telexes and seeing my public appearances.

…

Because we traded a number of different asset classes, and within each one we had programmed and tested lots of fundamental trading rules, we had many more high-quality ones to choose from than a typical manager who was tracking a smaller number of assets and was probably not trading systematically. I worked with Bob and Dan to pull our best decision rules from the pile. Once we had them, we back-tested them over long time frames, using the systems to simulate how the decision rules would have worked together in the past. We were startled by the results. On paper, this new approach improved our returns by a factor of three to five times per unit of risk, and we could calibrate the amount of return we wanted based on the amount of risk we could tolerate.

…

We quickly discovered that if we just tweaked what we did and created a new fund that managed money the same way as Pure Alpha but invested it solely in the most liquid markets, our expected returns would be the same and the expected risk (i.e., volatility) only slightly higher. We programmed this new approach into our computers, back-tested it to see how it worked in all countries and time frames, and explained it to our clients in detail so they could thoroughly understand the logic behind it. As much as I love and have benefited from artificial intelligence, I believe that only people can discover such things and then program computers to do them.

pages: 312 words: 91,538

The Fear Index
by Robert Harris
Published 14 Aug 2011

We’re still researching all these various categories to determine market impact and refine our model.’ Easterbrook raised his hand. ‘Yes, Bill?’ ‘Is this algorithm already operational?’ ‘Why don’t I let Hugo answer that, as it’s practical rather than theoretical?’ Quarry said, ‘Incubation started back-testing VIXAL-1 almost two years ago, although naturally that was just a simulation, without any actual exposure to the market. We went live with VIXAL-2 in May 2009, with play money of one hundred million dollars. When we overcame the early teething problems we moved on to VIXAL-3 in November and gave it access to one billion.

…

Then Quarry had hit the road of investors’ conferences, moving from city to city in the US and across Europe, pulling his wheeled suitcase through fifty different airports. He had loved this part – loved being a salesman, he who travels alone, walking in cold to an air-conditioned conference room in a strange hotel overlooking some sweltering freeway and charming a sceptical audience. His method was to show them the independently back-tested results of Hoffmann’s algorithm and the mouth-watering projections of future returns, then break it to them that the fund was already closed: he had only fulfilled his engagement to speak in order to be polite but they didn’t need any more money, sorry. Afterwards the investors would come looking for him in the hotel bar; it worked nearly every time.

pages: 89 words: 29,198

The Big Secret for the Small Investor: A New Route to Long-Term Investment Success
by Joel Greenblatt
Published 11 Apr 2011

Although various methods of weighting portfolios by economic fundamentals have been around since the late 1980s, in recent years fundamental indexes researched and constructed by Research Affiliates and WisdomTree Investments have become some of the most popular and widely accessible. In fact, Rob Arnott and his team at Research Affiliates has backtested (and in recent years made available) an index that has outperformed the market-cap-weighted S&P 500 by approximately 2 percent per year since 1962 (before expenses).7 The Research Affiliates index (known as the FTSE RAFI 1000 Index) was constructed using a five-year average of company cash flows, sales, dividends, and book value.

pages: 752 words: 131,533

Python for Data Analysis
by Wes McKinney
Published 30 Dec 2011

First, I’ll load historical prices for a portfolio of financial and technology stocks: names = ['AAPL', 'GOOG', 'MSFT', 'DELL', 'GS', 'MS', 'BAC', 'C'] def get_px(stock, start, end): return web.get_data_yahoo(stock, start, end)['Adj Close'] px = DataFrame({n: get_px(n, '1/1/2009', '6/1/2012') for n in names}) We can easily plot the cumulative returns of each stock (see Figure 11-2): In [117]: px = px.asfreq('B').fillna(method='pad') In [118]: rets = px.pct_change() In [119]: ((1 + rets).cumprod() - 1).plot() For the portfolio construction, we’ll compute momentum over a certain lookback, then rank in descending order and standardize: def calc_mom(price, lookback, lag): mom_ret = price.shift(lag).pct_change(lookback) ranks = mom_ret.rank(axis=1, ascending=False) demeaned = ranks - ranks.mean(axis=1) return demeaned / demeaned.std(axis=1) With this transform function in hand, we can set up a strategy backtesting function that computes a portfolio for a particular lookback and holding period (days between trading), returning the overall Sharpe ratio: compound = lambda x : (1 + x).prod() - 1 daily_sr = lambda x: x.mean() / x.std() def strat_sr(prices, lb, hold): # Compute portfolio weights freq = '%dB' % hold port = calc_mom(prices, lb, lag=1) daily_rets = prices.pct_change() # Compute portfolio returns port = port.shift(1).resample(freq, how='first') returns = daily_rets.resample(freq, how=compound) port_rets = (port * returns).sum(axis=1) return daily_sr(port_rets) * np.sqrt(252 / hold) Figure 11-2.

…

Cumulative returns for each of the stocks When called with the prices and a parameter combination, this function returns a scalar value: In [122]: strat_sr(px, 70, 30) Out[122]: 0.27421582756800583 From there, you can evaluate the strat_sr function over a grid of parameters, storing them as you go in a defaultdict and finally putting the results in a DataFrame: from collections import defaultdict lookbacks = range(20, 90, 5) holdings = range(20, 90, 5) dd = defaultdict(dict) for lb in lookbacks: for hold in holdings: dd[lb][hold] = strat_sr(px, lb, hold) ddf = DataFrame(dd) ddf.index.name = 'Holding Period' ddf.columns.name = 'Lookback Period' To visualize the results and get an idea of what’s going on, here is a function that uses matplotlib to produce a heatmap with some adornments: import matplotlib.pyplot as plt def heatmap(df, cmap=plt.cm.gray_r): fig = plt.figure() ax = fig.add_subplot(111) axim = ax.imshow(df.values, cmap=cmap, interpolation='nearest') ax.set_xlabel(df.columns.name) ax.set_xticks(np.arange(len(df.columns))) ax.set_xticklabels(list(df.columns)) ax.set_ylabel(df.index.name) ax.set_yticks(np.arange(len(df.index))) ax.set_yticklabels(list(df.index)) plt.colorbar(axim) Calling this function on the backtest results, we get Figure 11-3: In [125]: heatmap(ddf) Figure 11-3. Heatmap of momentum strategy Sharpe ratio (higher is better) over various lookbacks and holding periods Future Contract Rolling A future is an ubiquitous form of derivative contract; it is an agreement to take delivery of a certain asset (such as oil, gold, or shares of the FTSE 100 index) on a particular date.

Stocks for the Long Run, 4th Edition: The Definitive Guide to Financial Market Returns & Long Term Investment Strategies
by Jeremy J. Siegel
Published 18 Dec 2007

262 Index Options 264 Buying Index Options 266 Selling Index Options 267 The Importance of Indexed Products 267 Chapter 16 Market Volatility 269 The Stock Market Crash of October 1987 271 The Causes of the October 1987 Crash 273 Exchange-Rate Policies 274 The Futures Market 275 Circuit Breakers 276 The Nature of Market Volatility 277 Historical Trends of Stock Volatility 278 The Volatility Index (VIX) 281 Recent Low Volatility 283 The Distribution of Large Daily Changes 283 The Economics of Market Volatility 285 The Significance of Market Volatility 286 Chapter 17 Technical Analysis and Investing with the Trend 289 The Nature of Technical Analysis 289 Charles Dow, Technical Analyst 290 The Randomness of Stock Prices 291 Simulations of Random Stock Prices 292 Trending Markets and Price Reversals 294 Moving Averages 295 Testing the Dow Jones Moving-Average Strategy 296 Back-Testing the 200-Day Moving Average 297 The Nasdaq Moving-Average Strategy 300 CONTENTS CONTENTS xiii Distribution of Gains and Losses 301 Momentum Investing 302 Conclusion 303 Chapter 18 Calendar Anomalies 305 Seasonal Anomalies 306 The January Effect 306 Causes of the January Effect 309 The January Effect Weakened in Recent Years 310 Large Monthly Returns 311 The September Effect 311 Other Seasonal Returns 315 Day-of-the-Week Effects 316 What’s an Investor to Do?

…

.: Dow Jones-Irwin, 1988. 11 Historically, the daily high and low levels of stock averages were calculated on the basis of the highest or lowest price of each stock reached at any time during the day. This is called the theoretical high or low. The actual high is the highest level reached at any given time by the stocks in the average. CHAPTER 17 Technical Analysis and Investing with the Trend 297 Back-Testing the 200-Day Moving Average In Figure 17-2 are the daily and 200-day moving averages of the Dow Jones Industrial Average during two select periods: from 1924 to 1936 and 1999 to 2006. The time periods when investors are out of the stock market are shaded; otherwise, investors are fully invested in stocks.

Mastering Private Equity
by Zeisberger, Claudia,Prahl, Michael,White, Bowen , Michael Prahl and Bowen White
Published 15 Jun 2017

Notes 1 Please refer to Chapter 17 Fundraising for further details. 2 PE returns were calculated based on the INSEAD-Pevara dataset of 3,000 funds using quarterly modified internal rates of return. (http://centres.insead.edu/global-private-equity-initiative/research-publications/private-equity-navigator.cfm). Source for the MSCI World Index: Bloomberg. 3 Investors typically run extensive back-tests to explore how any new asset class might have impacted the portfolio performance historically. 4 Please refer to Chapter 24 Private Equity Secondaries for additional detail on the mechanics of secondaries. 5 Please refer to Chapter 1 Private Equity Essentials for further details on the J-curve. 6 The introduction of Basel III, the Dodd–Frank Act and the Volcker Rule for banks and Solvency II for insurance companies (in Europe) increased the cost of owning stakes in alternative investments. 7 Please refer to Chapter 19 Performance Reporting for a detailed description of the comparison of PE performance with other asset classes. 8 Refer to Chapter 18 LP Portfolio Management for a full overview of how LPs construct a portfolio. 9 Source: Greenhill Cogent 10 The number refers to professional fund management firms; but estimates vary widely on the number of active fund managers (source: Preqin 2016). 11 Before 2008, underlying companies were valued at cost until realization at exit, meaning that valuations of funds showed no volatility prior to the sale of the portfolio companies.

…

Also referred to as the articles of incorporation, certificate of incorporation, and other names in different jurisdictions. As-converted Basis A metric to determine the total equity base by assuming that all preferred shares have been converted into common shares based on a prespecified conversion ratio. Auction A sales process involving multiple competing parties to maximize the price for the seller. Back-testing Process of applying an allocation or trading strategy to historical data to gauge the effect on portfolio or investment performance. Bankruptcy Legal status in which an insolvent company (which cannot fully repay its debt) is declared bankrupt, typically by court order. Base Case Financial Scenario Scenario based on the company's expected/most likely operating performance.

pages: 172 words: 49,890

The Dhandho Investor: The Low-Risk Value Method to High Returns
by Mohnish Pabrai
Published 17 May 2009

Greenblatt goes on to suggest that an investor ought to build a portfolio of about 25 to 30 of these Magic Formula stocks. He recommends buying five to seven of them every two to three months. After a given stock has been held for a year, it is sold and replaced with another one from the updated Magic Formula list. Greenblatt’s back-testing showed that these Magic Formula stocks have generated returns as high as 20 percent to 30 percent annualized. It trounces the S&P 500—with no thinking or analysis required. If we step back and thinks about it, the Magic Formula is effectively an index. But it is the mother of all indexes—an index on steroids.

pages: 161 words: 51,919

What's Your Future Worth?: Using Present Value to Make Better Decisions
by Peter Neuwirth
Published 2 Mar 2015

Taleb illustrates the above point by making the distinction between Newtonian physics, which was disproved by experiments that led to Einstein’s Theory of Relativity, and the pseudoscience of Astrology where the theory can always be tweaked and expanded to explain any observed behavior or personality characteristic that doesn’t appear to come directly from an individual’s initial chart. I agree with Taleb that econometrics is much more like Astrology than it is like Physics.28 But even worse, this ability to continuously “backtest” and recalibrate one’s models has led vast numbers of otherwise rational people to believe that the econometric models informing The Wall Street Journal pronouncements by “leading economists” regarding the likely future course of the economy are more valid than the daily horoscope, which they dismiss as nonsense.

pages: 356 words: 51,419

The Little Book of Common Sense Investing: The Only Way to Guarantee Your Fair Share of Stock Market Returns
by John C. Bogle
Published 1 Jan 2007

Bogle is against the widespread practice today of building portfolios that consist of 10–15 asset classes, whose sole purpose is to create complexity to generate fees for greedy asset managers. Keeping it simple means owning stocks and some bonds. It also means not getting too fancy and too carried away by adding fashionable asset classes whose merits are derived from a backtested computer model.” Chapter Seven The Grand Illusion Surprise! The Returns Reported by Mutual Funds Are Rarely Earned by Mutual Fund Investors. IT IS GRATIFYING THAT industry insiders such as Fidelity’s Peter Lynch, former Investment Company Institute (ICI) chairman Jon Fossel, Mad Money’s James Cramer, and AQR’s Clifford Asness agree with me, as you may recall from Chapter 4.

pages: 848 words: 227,015

On the Edge: The Art of Risking Everything
by Nate Silver
Published 12 Aug 2024

You get to the long run much faster than in something like the stock market, where corporate strategies can take years to play out. So if sportsbooks are posting betting lines that can be beat by a college senior, then perhaps markets are not as efficient as economists say. Still, it’s one thing to claim you’ve found a profitable betting opportunity when you’re backtesting a statistical model. It’s another thing to actually bet on it—to have some skin in the game—and win. “If your model is so smart, why don’t you bet on it?” is a common refrain in the River. Sometimes there are good reasons not to.[*6] As I hope you’ll see from this chapter, it is far from trivial to get your money down even if you theoretically have a profitable bet.

…

For instance, the tokens “coyote” and “roadrunner” would be closely linked and would draw more attention from the model. See also: transformer. Backdoor: Salvaging a losing bet through unlikely means, e.g., by catching two consecutive cards to make a flush, or when a team kicks a meaningless field goal to beat the point spread; the karmic opposite of a bad beat. Backtest: Evaluating a model’s performance on known, past data. Less reliable than using out-of-sample data, where you don’t know the outcomes ahead of time. See also: overfitting. Bad beat: Losing a bet that you were a clear favorite to win, especially if your opponent made a speculative play. A♣A♦ losing to 7♥2♣ (one of the worst hands in poker) is considered more of a bad beat than aces losing to another strong hand like pocket kings.

…

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z A abstraction, 23–24, 29, 30–31, 130, 477 academia, 26, 27, 28, 294–96 See also Village accelerationists, 31, 250, 411–13, 455–56, 477, 539n accelerators, 405–6, 477 action, 477 adaptability, 235–37, 264 addiction, 164–65, 166, 167, 168–69, 213–14, 321 Addiction by Design: Machine Gambling in Las Vegas (Schüll), 154–55 Adelson, Sheldon, 146 Adelstein, Garrett, 100, 102, 106 Robbi hand, 80–86, 89, 117, 123–29, 130, 444–45, 512n advantage play, 158–61, 478 adverse selection, 478 Aella, 375–77 Age of Em, The (Hanson), 379 agency, 453, 469–70, 478 agents, 478 aggressiveness, 120 AGI (artificial general intelligence), defined, 478 Aguiar, Jon, 199 AI (artificial intelligence) accelerationists, 31, 250, 411–13, 455–56, 477, 539n adaptability and, 236n agency and, 469–70 alignment and, 441–42, 478 Sam Altman and, 406 analogies for, 446, 541n bias and, 440n breakthrough in, 414–15 commercial applications, 452–53 culture wars and, 273 decels, 477 defined, 478 economic growth and, 407n, 463–64 effective altruism and, 21, 344, 348, 355, 359, 380 engineers and, 411–12 excitement about, 409–10 impartiality and, 359, 366 moral hazard and, 261 New York Times lawsuit, 27, 295 OpenAI founding, 406–7, 414 optimism and, 407–8, 413 poker and, 40, 46–48, 60–61, 430–33, 437, 439, 507n poor interpretability of, 433–34, 437, 479 prediction markets and, 369, 372 probabilistic thinking and, 439 randomization and, 438 rationalism and, 353, 355 regulation of, 270, 458, 541n religion and, 434 risk impact and, 91 risk tolerance and, 408 River-Village conflict and, 27 SBF and, 401, 402 sports betting and, 175–76 technological singularities and, 449–50, 497 transformers, 414–15, 434–41, 479, 499 Turing test and, 499 See also AI existential risk AI existential risk accelerationists and, 412–13, 455–56, 539n alignment and, 441 arguments against, 458–60 Bid-Ask spread and, 444–46 commercial applications and, 452–53 Cromwell’s law and, 415–16 determinism and, 297 effective altruism/rationalism and, 21, 355, 456 EV maximizing and, 457 excitement about AI and, 410 expert statement on, 409, 539n Hyper-Commodified Casino Capitalism and, 452–53 instrumental convergence and, 418 interpretability and, 433–34 Kelly criterion and, 408–9 models and, 446–48 Musk and, 406n, 416 optimism and, 413–14 orthogonality thesis and, 418 politics and, 458, 541n prisoner’s dilemma and, 417 reference classes and, 448, 450, 452, 457 societal institutions and, 250, 456–57 takeoff speed and, 418–19, 498 technological Richter scale and, 450–52, 451, 498 Yudkowsky on, 372, 415–19, 433, 442, 443, 446 Alexander, Scott, 353, 354, 355, 376–77, 378 algorithms, 47, 478 alignment (AI), 441–42, 478 all-in (poker), 478 alpha, 241–42, 478 AlphaGo, 176 Altman, Sam, 401 AI breakthrough and, 415 AI existential risk and, 419n, 451, 459 OpenAI founding and, 406–7 OpenAI’s attempt to fire, 408, 411, 452n optimism and, 407–8, 413, 414 Y Combinator and, 405–6 Always Coming Home (Le Guin), 454–55, 541n American odds, 477, 491 American Revolution, 461 analysis, 23, 24, 478 analytics casinos and, 153–54 defined, 23, 478 empathy and, 224 limitations of, 253–54, 259 politics and, 254 sports betting and, 171, 191 venture capital and, 249 anchoring bias, 222n, 478 Anderson, Dave, 219–20, 230, 231 Andreessen, Marc accelerationists and, 411 AI analogies and, 446, 541n AI existential risk and, 446 Adam Neumann and, 281–82 on patience, 260 politics and, 267–68 River-Village conflict and, 295 techno-optimism and, 249, 250–51, 270, 296, 498 VC profitability and, 293, 526n VC stickiness and, 290, 291–92 angles, 192–94, 235–36, 305, 478 angle-shooters, 478 ante (poker), 478 anti-authority attitude, 111–12, 118, 137 See also contrarianism apeing, 479 arbitrage (arb), 171, 172–74, 206, 478, 489, 516n, 517n Archilochus, 236, 263, 485 Archipelago, The, 22, 310, 478 arms race, 478 See also mutually assured destruction; nuclear existential risk art world, 329–30, 331n ASI (artificial superintelligence), 478 Asian Americans, 135–36, 513n See also race asymmetric odds, 248–49, 255, 259, 260–62, 276, 277 attack surfaces, 177, 187, 478 attention (AI), 479 attention to detail, 233–35 autism, 282–84, 363, 525n B back doors, 479 backtesting, 479 bad beats, 479 “bag of numbers,” 433, 479 bank bailouts, 261 Bankman, Joseph, 383–84 Bankman-Fried, Sam (SBF) AI and, 401, 402 angles and, 305 attitude toward risk, 334–35 bankruptcy and arrest of, 298–301, 373–74 cryptocurrency business model and, 308–9 cults of personality and, 31, 338–39 culture wars and, 341n as dangerous, 403–4 disagreeability and, 280 effective altruism and, 20, 340–42, 343, 374, 397–98, 401 as focal point, 334 fraud and, 124, 374 Kelly criterion and, 397–98 moral hazard and, 261 NOT INVESTMENT ADVICE and, 491 personas of, 302 politics and, 26, 341n, 342 public image of, 338 responses to bankruptcy and arrest, 303–5, 383–85, 386–88 risk tolerance and, 334–35, 397–403, 537–38n River and, 299 theories of, 388–96 trial of, 382–83, 385–86, 387, 403 utilitarianism and, 360, 400, 402–3, 471, 498 venture capital and, 337–39 warning signs, 374 bankrolls, 479 Baron-Cohen, Simon, 101n, 283, 284 Barzun, Jacques, 466 baseball, 58–59, 174 See also sports betting base rates, 479 basis points (bips), 479 basketball, 174 See also sports betting Bayesian reasoning, 237, 238, 353, 355, 478, 479, 493–94, 499 Bayes’ theorem, 479 beards, 207–8, 479 See also whales bednets, 479 Bennett, Chris, 177, 178 Bernoulli, Nicolaus, 498 Betancourt, Johnny, 332–33 bet sizing, 396, 479 Bezos, Jeff, 277, 410 Bid-ask spread, 444–46, 479 Biden, Joe, 269, 375 big data, 432–33, 479 Billions, 112 Bitcoin bubble in, 6, 306, 307, 307, 310, 312 creation of, 322–23, 496 vs.

pages: 178 words: 52,637

Quality Investing: Owning the Best Companies for the Long Term
by Torkell T. Eide , Lawrence A. Cunningham and Patrick Hargreaves
Published 5 Jan 2016

Cristina Cruz Serrano and Laura Nuñez Letamendia, ‘Value Creation in Listed European family firms (2001-2010)’, Academy Management Journal (2015). ‘The unsung masters of the oil industry’, The Economist, June 2012. Safilo relisted in December 2005 at over €60/ share, since which point its share price has declined over 80% in absolute terms. Back-testing Credit Suisse’s HOLT data, we found that a strategy of owning eCAP stocks (eCAPs refer to stocks which have sustained superior CFROI over five or more years) would have delivered outperformance of the Global market in 15 out of the last 20 years. This is broadly consistent with the findings of the Goldman Sachs SUSTAIN team, which see an even more consistent long-term pattern of outperformance by first quartile CROCCI companies on a sector relative basis.

pages: 542 words: 145,022

In Pursuit of the Perfect Portfolio: The Stories, Voices, and Key Insights of the Pioneers Who Shaped the Way We Invest
by Andrew W. Lo and Stephen R. Foerster
Published 16 Aug 2021

The call option depends on five factors: the price of the underlying security (here, the price level of the S&P 500 index), the exercise price, the time to maturity or expiration of the option (again, we’ll assume three months), the risk-free interest rate (the three-month Treasury bill rate), and the volatility of the underlying security or, in this case, the volatility of a proxy for the U.S. equity market. We can readily obtain the first four factors. The only one not directly observable is the volatility. FIGURE 6.1: The CBOE Volatility Index, or VIX, January 1986 to July 2020 (back-tested prior to 2004). Source: “VIX Index Historical Data,” CBOE, http://www.cboe.com/products/vix-index-volatility/vix-options-and-futures/vix-index/vix-historical-data. Let’s assume that traders have properly priced this particular option. If that’s the case, then we can back out the implied volatility of the market, since we know the other four factors and the price.

…

According to Black (1989a), there were fewer opportunities to make money from mispriced options by the late 1980s compared to the early 1970s, because by then traders used the Black-Scholes/Merton option pricing model extensively. 65. See “OTC Derivatives Statistics at End-December 2019,” Bank for International Settlements, May 7, 2020, https://www.bis.org/publ/otc_hy2005.htm. 66. Trading on the VIX index commenced in 2004. The levels back to 1986 are based on CBOE back-testing, showing what the level would have been given S&P 500 option prices at the time. While the official record closing high on the VIX of 82.69 occurred on March 16, 2020, near the start of the COVID-19 pandemic, the October 19, 1987, level would have been almost twice as high, at 150.19. 67. Interview with authors. 68.

pages: 244 words: 58,247

The Gone Fishin' Portfolio: Get Wise, Get Wealthy...and Get on With Your Life
by Alexander Green
Published 15 Sep 2008

(Also, the Vanguard Inflation-Protected Securities Fund wasn’t launched until June 29, 2000. In the hypothetical example above, I substituted the performance of TIPS themselves for the three years prior to the Vanguard fund’s debut.) Of course, you cannot know for certain what this portfolio (or any other) will return in the future, no matter how much back-testing you do. As I’ve said from the beginning, to a large extent uncertainty will always be your companion. Historical asset class returns are a valuable guidepost. They are not a guarantee. However, it’s not just the attractiveness of past or future returns that make the Gone Fishin’ Portfolio such a compelling choice.

pages: 224 words: 13,238

Electronic and Algorithmic Trading Technology: The Complete Guide
by Kendall Kim
Published 31 May 2007

Factor analyst This stock selection research service leverages over 300 different stock selection indicators maintained and updated for portfolio construction and stock selection. . Virtual research analyst Portfolio managers can use this service to support any disciplined stock selection strategy. This research enables customization of candidate identification criteria, quick screening, backtesting, and quality control. Profiling the Leading Vendors 171 . T-Cost Pro A Web-based transaction cost management product capable of producing detailed analysis of time-stamped executions on a Tþ1 basis. QSG products are designed to help buy-side firms overcome the mediocrity associated with using simple benchmarks such as VWAP to conduct transaction cost analysis.

pages: 236 words: 77,735

Rigged Money: Beating Wall Street at Its Own Game
by Lee Munson
Published 6 Dec 2011

If it were 2010 he would be living at home begging you for the money. Only this time around, taxes can be avoided by investing in an IRA. By 1990, the wind is really at his back, because trades are cheaper from discount brokerage firms that didn’t previously exist, he can defer taxes because IRAs existed after 1974, and it’s easier to get the information to back-test investments, even before the Internet was widely used. 1990–2010 Results In 1990, Cornelius calculates that with his $100 he can buy 2.76 shares of Disney, 1.23 shares of Eastman Kodak, 1.18 shares of IBM, 3.45 shares of Coca-Cola, and 19.23 shares of Philip Morris (Table 1.3). Just like you, his broker doesn’t charge him any commissions.

pages: 192 words: 75,440

Getting a Job in Hedge Funds: An Inside Look at How Funds Hire
by Adam Zoia and Aaron Finkel
Published 8 Feb 2008

Search 1: Hedge Fund Risk Analyst Note: This fund has a director of risk management who is looking for an additional resource (risk analyst) to join his team and develop within the firm. Description • Responsible for periodic report production, including: • Value at risk (VaR) and volatility reporting by portfolio. • Back-testing and historical performance measurement. • Portfolio segmentation analysis. • Factor analysis reporting. • Position level: • Expected return by position. • Risk analysis by position. • Marginal impact. • Relative risk/reward performance: • Stress testing. c07.indd 91 1/10/08 11:08:07 AM 92 Getting a Job in Hedge Funds • Correlation and concentration reporting by name, sector, and industry. • Responsible for the development and maintenance of a risk management database: • Creation of a centralized risk management database repository. • Daily data extraction from trading systems (Eze Castle) and accounting systems (VPM). • Maintenance of a security master and entity master tables. • Sourcing and storage of market pricing information. • Data cleaning and standardization. • Automation of data feeds from the risk management database to other applications (e.g., RiskMetrics) or models. • Supporting portfolio analysis: • Position and portfolio volatility analysis. • Correlation and factor model development. • Relative risk-adjusted performance measurement. • Historical and prospective analysis. • Analysis by position, portfolio, strategy, and so on. • Ad hoc analysis of portfolio.

pages: 249 words: 77,342

The Behavioral Investor
by Daniel Crosby
Published 15 Feb 2018

And the size of the outperformance was sizeable – 1% per month, even after adjusting for return differences owing to other risk factors.127 Indeed, the effects of momentum tend to be pervasive and not limited with respect to market, place or time. Chris Geczy and Mikhail Samonov conducted what is affectionately referred to as the “world’s longest backtest” and found that momentum effects have persisted in the US since 1801!128 Momentum signals have worked well in the UK since the Victorian Age (Chabot, Ghysels and Jagannathan, 2009) and have proven their power and persistence across 40 countries and more than a dozen asset classes.129 So deep-seated are our psychological tendencies toward momentum that, “the momentum premium has been a part of markets since their very existence, well before researchers studied them as a science.”

pages: 317 words: 84,400

Automate This: How Algorithms Came to Rule Our World
by Christopher Steiner
Published 29 Aug 2012

Peterffy needed a way to express all of this in one elegant algorithm that rightly weighted each factor. It was a complicated math problem that he found nearly impossible to solve. He cycled through spurts of dejection and inspiration. After working on the problem for more than a year, Peterffy devised an algorithm of differential equations that cleverly weighted all of the ingredients. He back-tested the algorithm to see if it would have made money in the past, but the data sets for commodities options at that point in history were limited. This was before computers handled such things adeptly and, more important, before the options market had much history. So Mocatta did the only thing it could: it started trading with the algorithm.

pages: 302 words: 86,614

The Alpha Masters: Unlocking the Genius of the World's Top Hedge Funds
by Maneet Ahuja , Myron Scholes and Mohamed El-Erian
Published 29 May 2012

“It is believing that things that worked in the past will continue to work and leveraging up to be on it. Nowadays, with the computer, it is easy to identify what would have worked and, with financial engineering, to create overoptimized strategies. I believe we are entering a period that will not be consistent with the back-testing, and problems will arise. When that dynamic exists and there’s close to zero interest rate, we knew that the ability of the central bank to ease monetary policy is limited.” When Dalio looks at the world today, he sees it divided into two parts—debtor-developed deficit countries and emerging market creditor countries.

pages: 394 words: 85,252

The New Sell and Sell Short: How to Take Profits, Cut Losses, and Benefit From Price Declines
by Alexander Elder
Published 1 Jan 2008

Riding a trend is like riding a bucking horse that tries to shake you off. Holding on to a trend-following trade requires a great deal of patience and self-assurance—a lot of mental work. Question 7—System vs. Discretionary Trading Answer 3 Selection “A greater degree of freedom” is incorrect. System traders who have done a lot of backtesting can have a fairly high level of confidence knowing what profits or losses to expect down the road. If they have the discipline to follow all the signals of their system, they will lower their stress level, insulating themselves to a degree from uncertainty in the markets. What they give up is the freedom to make decisions as market conditions change, creating new threats or opportunities.

pages: 268 words: 81,811

Flash Crash: A Trading Savant, a Global Manhunt, and the Most Mysterious Market Crash in History
by Liam Vaughan
Published 11 May 2020

X Navinder Sarao’s arrest would ultimately involve an army of people from agencies including the CFTC, the FBI, the Metropolitan Police, and the Department of Justice, but the investigation started with a lone individual with no affiliation to the government. He was a day trader, like Sarao, grinding out a living in a small prop firm in Chicago who, in 2012, happened to be back-testing his system using data from the day of the Flash Crash when he spotted something the whole world had missed. His identity has never been made public. We’ll call him Mr. X. Mr. X is a few years older than Sarao and started his career as part of a pit trading firm’s first forays into electronic trading.

pages: 303 words: 84,023

Heads I Win, Tails I Win
by Spencer Jakab
Published 21 Jun 2016

Or I could have a service tell me what to do. I tried one out called MyPlanIQ. Their free service asked me for my risk profile (moderate) but not any other information and suggested a portfolio of five index funds available in my company’s plan that it would tell me by e-mail to rebalance quarterly. I “backtested” the results and found that the returns over ten years would have been 26 percent better than the higher-cost Fidelity 2035 fund. An enhanced version of the service that claims to have a more optimized formula costs $200 a year. Human financial advisors are more expensive and may or may not assist you with a 401(k) plan since it’s not under their control.

pages: 345 words: 87,745

The Power of Passive Investing: More Wealth With Less Work
by Richard A. Ferri
Published 4 Nov 2010

What qualifies as an index has broadened over the years as more ETFs come to market that follow highly customized nonstandard index methods. Today, it seems as though anything can be called an index. An index provider merely creates a mechanical set of rules for security selection, security weighting, and trading, and publishes their back-tested results. For example, an index may be made up of only dividend paying stocks with those stocks being weighted by dividend yield. Or, an index could include companies located west of the Mississippi that have female CEOs under the age of 50. Such an index doesn’t exist, but it would if a fund company thought they could sell an index fund or ETF to enough people based on that index.

Deep Value
by Tobias E. Carlisle
Published 19 Aug 2014

First and foremost, I’d like to thank my wife, Nickole, who took over the primary parental responsibilities for our newborn, Auristella, whose arrival marked the midpoint of the preparation of the first draft. I’d like to thank the early reviewers of that primordial first draft: Scott Reardon, Taylor Conant, Travis Dirks, PhD, Peter Love, Toby Shute, and my mother and father, Drs. Wendy and Roger Carlisle. I’d like to thank Jeffrey Oxman, PhD for his assistance with backtesting the various strategies discussed in the book. Finally, I appreciate the assistance of the team at Wiley Finance, most especially Bill Falloon, Lia Ottaviano, Angela Urquhart, Tiffany Charbonier, and Meg Freeborn, who provided guidance and advice along the way. xv About the Author Tobias Carlisle is the founder and managing director of Eyquem Investment Management LLC, and serves as portfolio manager of Eyquem Fund LP.

pages: 267 words: 90,353

Private Equity: A Memoir
by Carrie Sun
Published 13 Feb 2024

I did have to answer some very broad questions about myself: “Why are you here?” “Why do you want this job?” Boone asked me to explain what I had done at Fidelity; I perked up. I told him how, with no real programming experience, I had learned Java and R and built, from scratch, a custom portfolio optimizer that we used for our quant-investment needs: back-tests, simulations, rebalancings. “Basically an in-house version of Barra,” I said. What’s Barra? he asked. It was then I realized how wide the knowledge gap was between fundamentally and quantitatively driven finance: here was an investor whom many had called a phenom, who did not know one of the industry-leading providers of optimizers and risk models.

Work Less, Live More: The Way to Semi-Retirement
by Robert Clyatt
Published 28 Sep 2007

Using data going back to the 1870s for 50-year payout periods, Greaney finds optimal stock/bond allocation at around 80% stocks, with a resulting 3.35% Safe Withdrawal Rate, inflation-adjusted annually. Studies results are primarily focused around establishing a 100% safe rate. The study affirms approximately 90% safety around the 4% withdrawal rate level. 352 | Work Less, Live More Guyton. Jonathan T. Guyton develops and back-tests various additional withdrawal rules that would have helped permit higher withdrawal rates while maintaining portfolio safety over time. Specifically, he develops rules for limiting inflation adjustments following “down” years and limiting any inflation adjustments to 6%. Problems would arise with these rules, however, since there is no “catch-up” in future years—a down year or inflation cap permanently lowered the real value of withdrawals for the lifetime of the semi-retiree.

pages: 317 words: 106,130

The New Science of Asset Allocation: Risk Management in a Multi-Asset World
by Thomas Schneeweis , Garry B. Crowder and Hossein Kazemi
Published 8 Mar 2010

It is rather like the story of the individual asking for help in finding his watch only to Other Liquidity risk Regulatory risk Concentration risk Leverage risk Operational risk Key person Reputational risk Counterparty risk Transparency risk Model risk Complexity risk Derivatives risk Credit risk Market risk Written Due Diligence Valuation Policies Stress Testing Performance Measures Written Policies and Guidelines Acknowledged Fiduciaries Adequate Systems and Procedures Risk Limits Model Review Backtesting Independent Risk Oversight Backup and Disaster Recovery Education and Knowledge Clear Organization Structure Compliance monitoring EXHIBIT 2.1 Array of Risk Determinants 23 24 THE NEW SCIENCE OF ASSET ALLOCATION be asked: Where did you lose it? When responding that he lost it across the street but is looking under the lamp because the light is better here directly illustrates the point.

pages: 289 words: 95,046

Chaos Kings: How Wall Street Traders Make Billions in the New Age of Crisis
by Scott Patterson
Published 5 Jun 2023

He didn’t do Twitter, and he certainly didn’t get into public sparring matches with other hedge fund managers. But he shared Taleb’s contempt for AQR’s tail-hedging research. Was it a surprise AQR couldn’t replicate Universa’s returns? He’d been perfecting the strategy for more than two decades. “Universa’s 12 years of results speak for themselves,” Spitznagel told Institutional Investor. “Backtests from naïve researchers are an unfortunate distraction.” What angered Spitznagel most of all was that AQR’s research played a role in convincing CalPERS to throw away the windfall it could have made in March 2020. And that all the money it could have made in the coming years for its pensioners was up in smoke.

pages: 416 words: 118,592

A Random Walk Down Wall Street: The Time-Tested Strategy for Successful Investing
by Burton G. Malkiel
Published 10 Jan 2011

Real money investment strategies don’t produce the results that academic papers say they should. Roll’s final point was underscored for me during an exchange I had with a portfolio manager who used the most modern quantitative methods to run his portfolio on the basis of all the statistical work done by academics and practitioners. He “back-tested” his technique with historical data over a twenty-year period and found that it outperformed the Standard & Poor’s 500-Stock Index by three percentage points per year. But when he started using those quantitative methods with real money, his results were quite different. Over the next twenty-year period, he barely managed to equal the S&P return after expenses.

pages: 289 words: 113,211

A Demon of Our Own Design: Markets, Hedge Funds, and the Perils of Financial Innovation
by Richard Bookstaber
Published 5 Apr 2007

In mid-2002 the performance of stat arb strategies began to wane, and the standard methods have not recovered. This is not surprising, given the simplicity of the strategies, the ease of entry, and the proliferation of computer power. My son David had the bad luck to get started in this sort of strategy just as the window of opportunity was closing. The strategy had performed admirably in years of back-tests and in the first months of operation, but then sputtered along doing next to nothing. He closed it down from active trading after six months and then ran it on paper for another year, with no better results. The stat arb concept remains, but in place of the stat arb strategies of the late 1980s and the 1990s is an incarnation called high frequency trading.

pages: 425 words: 122,223

Capital Ideas: The Improbable Origins of Modern Wall Street
by Peter L. Bernstein
Published 19 Jun 2005

Although Fama’s efforts to develop profitable trading rules were by no means unsuccessful, the ones he found worked only on the old data, not on the new. He did not realize it at the time, but his frustrating experience was shared by many highly motivated investors seeking ways to beat the market. All too often, backtests give every promise of success but prove disappointing when investors try to apply them in real time. The environment shifts, market responses slow down or speed up, or too many people follow the same strategy and end up competing away one another’s potential profits. Like Alfred Cowles before him, Fama grew curious about why ideas that seem good on paper produce such disappointing results when real money is riding on them.

pages: 412 words: 122,655

The Fund: Ray Dalio, Bridgewater Associates, and the Unraveling of a Wall Street Legend
by Rob Copeland
Published 7 Nov 2023

The money was simply the proof for what would become a crucial logic chain: Bridgewater’s investment rules had helped produce Dalio’s wealth, thus following those rules could be said to lead to wealth. Those investment rules were culled from Bridgewater’s economic research, which Dalio credited to his tendency to back-test, challenge, and amend as needed. Thus the money proved to Dalio that his philosophical approach worked. One could not exist without the other. Like technology titans who clothed themselves in a higher purpose than merely slinging online goods or apps, Dalio’s life work began to seem more important than the running of an investment firm.

pages: 537 words: 144,318

The Invisible Hands: Top Hedge Fund Traders on Bubbles, Crashes, and Real Money
by Steven Drobny
Published 18 Mar 2010

We then conducted a simple study that adds leveraged bond positions to a portfolio of 100 percent long domestic U.S. equities. The back test results, from 1992 to 2009, show that adding 100 percent leverage to buy U.S. Treasuries increased annual yield by almost 5 percent while reducing the worst drawdown by 10 percent. Back-testing other periods, such as 1940 to 1980, yield less conclusive results; but it is clear more analytical work needs to be done in this area. It is also much too facile to say that leverage is bad in every occasion. Logically, since bonds can be repo’d at the cash rate and have a risk premium over cash, over time the cost of such insurance should actually be a positive to the fund (see box).

pages: 431 words: 132,416

No One Would Listen: A True Financial Thriller
by Harry Markopolos
Published 1 Mar 2010

It took several months of playing with numbers to fulfill those parameters. Neil, of course, was a major contributor, and I got a lot of data from various major firms. From Citigroup for example, I got the complete S&P 500 price return histories from 1926 to the day I received it. Then I began putting things in, taking things out, testing and retesting and back-testing to see how each package would perform in various market environments. I did this knowing full well that Bernie hadn’t bothered to do any of this. He just sat down and made it up. It’s considerably easier that way—and you always get the results you want! Eventually I developed a product we named the Rampart Options Statistical Arbitrage.

Commodity Trading Advisors: Risk, Performance Analysis, and Selection
by Greg N. Gregoriou , Vassilios Karavas , François-Serge Lhabitant and Fabrice Douglas Rouah
Published 23 Sep 2004

He has been involved in the commodity markets since 1994. Prior to joining Premia, he developed programmed trading applications for Morgan Stanley’s Equity Division and proprietary computer models for urban economics. From 1994 to 1998 he worked in the Derivative Strategies Group of Putnam Investments where he researched, back-tested, and implemented relative-value derivatives strategies. Mr. Eagleeye holds a degree in Applied Mathematics from Yale University and an M.B.A. from the University of California at Berkeley. Andrew Green graduated in March 2004 with an MBA degree in Finance from Thunderbird, the American Graduate School of International Management.

pages: 444 words: 151,136

Endless Money: The Moral Hazards of Socialism
by William Baker and Addison Wiggin
Published 2 Nov 2009

Siegel has escaped from the world of academia into the lucrative world of Wall Street through establishing WisdomTree, a provider of ETFs and mutual funds. WisdomTree tweaks the major indicies to squeeze out a slightly better return with less volatility—all based upon statistical analysis thoughtfully proven through roughly 40 years of backtesting. The strategy is to exploit a structural flaw that requires index funds to buy more of stocks that go up and sell as underperformers go down;—instead it does the opposite by slightly overweighting holdings of high dividend yielding or low P/E stocks. While the approach appears to be successful and probably improves upon the returns of individuals plunging their IRAs into hot tips heard at the country club tap room, demand for this product may be indicative of the public’s unwavering faith in equities and bonds, and buying-on-dips right up until the end.

Investment: A History
by Norton Reamer and Jesse Downing
Published 19 Feb 2016

The computer gave ﬁnancial practitioners access to a wealth of information and data that was previously rather intractable to synthesize and without which it was virtually impossible to test rigorous models. Further, the very notion 264 Investment: A History of a quantitative fund—a “quant” fund—or a quantitative strategy is simply inconceivable without the computer. Without the aid of the computer, one could not construct and back-test robust models or even generate signals where certain criteria were met. The Hedge Fund Universe Today As of 2014, the hedge fund industry had approximately $2.5 trillion in assets under management. Additionally, approximately $455 billion was in funds of hedge funds, a diversiﬁed investment vehicle designed to add value by selecting and overseeing other hedge fund managers that generate alpha.20 Before discussing these funds of funds in more detail, let us consider the various strategies of individual funds.

pages: 499 words: 148,160

Market Wizards: Interviews With Top Traders
by Jack D. Schwager
Published 7 Feb 2012

Do you use trading systems? No, we are “discretionary traders.” We only use technical indicators and systems as trading tools. One particularly interesting system we have developed is based on quirks related to volatility. Our belief is that volatility offers clues to trend direction. Although we’ve found through backtesting that this system gives you good signals, we do not blindly follow the trades. For an automated, computerized system, what would you consider good results in terms of average annual return? About 40 to 50 percent, with maximum equity drawdowns under 10 percent. But systems will always have larger maximum drawdowns than that.

Trade Your Way to Financial Freedom
by van K. Tharp
Published 1 Jan 1998

There are some stocks and commodities that do not trend very well and make poor trend-following candidates. Likewise, there are those that have ranges that are too tight for band trading or do not trade well in ranges (they frequently extend far past bands, for example). These can be identified only through experience and/or backtesting. Editor’s Comments Band trading typically gives you lots of trading opportunities, and it is excellent for short-term traders. Thus, if you like (1) lots of trading activity, (2) selling highs, and (3) buying lows, then some form of band trading might be right for you. If you look at the charts, you’ll see many examples that work very well and many examples that do not work at all.

pages: 566 words: 163,322

The Rise and Fall of Nations: Forces of Change in the Post-Crisis World
by Ruchir Sharma
Published 5 Jun 2016

The reason I developed rules at all was to focus my eyes and those of my team on what matters. When we visit a country, we gather impressions, storylines, facts, and data. While insight is embedded in all observations, we have to know which ones have a reliable history of telling us something about a nation’s future. The rules systematize our thoughts and have been back-tested to determine what has worked and what has not. Eliminating the inessential helps steer the conversation to what is relevant in evaluating whether a country is on the rise or in decline. I have narrowed the voluminous lists of growth factors to a number that is large enough to keep the most significant forces of change on our radar but small enough to be manageable.

pages: 1,239 words: 163,625

The Joys of Compounding: The Passionate Pursuit of Lifelong Learning, Revised and Updated
by Gautam Baid
Published 1 Jun 2020

The ability to keep investing at regular intervals, to stay the course through thick and thin, ups and downs, and bull markets and bear markets, and to not worry where the markets are going tomorrow, or next week, or next month is what matters. Simple. It’s simple but not easy. Emotions cannot be back-tested; that’s why all previous bear markets and the accompanying cheap stock valuations prevailing at those times look like cinch opportunities only in hindsight. From the time of its IPO in 1980 through 2012, Apple was a 225-bagger. A $10,000 investment in it turned into $2.25 million. But an investor would have had to suffer through two declines of more than 80 percent and several drops of more than 40 percent along the way.

pages: 733 words: 179,391

Adaptive Markets: Financial Evolution at the Speed of Thought
by Andrew W. Lo
Published 3 Apr 2017

Amir Khandani had just come back to MIT from a summer internship and was looking for a thesis topic. I suggested that we try to figure out what happened during the Quant Meltdown by simulating a simple quantitative equity trading strategy.26 A common practice in the investment business is to evaluate a particular strategy by performing a “backtest,” or “paper trading,” where you use historical prices to calculate the realized profits and losses of trades that the strategy would have called for. For example, suppose a superstitious friend tells you that you should never buy stocks on Friday the Thirteenth—is that good advice? One way to evaluate this advice is to compute the average return of the S&P 500 index between Fridays and Mondays for all Fridays that fall on the Thirteenth, and then do the same for all non-Thirteenth Fridays and Mondays, and compare the two averages.

pages: 1,042 words: 266,547

Security Analysis
by Benjamin Graham and David Dodd
Published 1 Jan 1962

(pp. 240–241) Today, of course, the activities of creditors’ committees, which play a major role in reorganizations, are closely supervised. 15 Neporent’s testimony is available at judiciary.senate.gov. 16 Bill Miller, “Good Times Are Coming!” Time, March 8, 2005. 17 Kenneth L. Fisher, 100 Minds That Made the Market (New York: Wiley, 2007), p. 61. Fisher goes on to observe of this late-in-life conversion: “Ironically, Graham’s adoption of ‘the efficient market’ was just before computer backtests would poke all kind of holes in that theory.” 18 Kenneth Lee, Trouncing the Dow: A Value-Based Method for Making Huge Profits (New York: McGraw-Hill, 1998), pp. 1–2. 19 In Berkshire Hathaway’s 2000 annual report, Buffett said of his experience in Graham’s class that “a few hours at the feet of the master proved far more valuable to me than had ten years of supposedly original thinking.” 20 Hamlet, Act III, Scene 2. 1 In the 1934 edition we had here a section on investment-quality senior issues obtainable at bargain levels.