Systematic Trading

Preamble Preface - Systematic trading and investing - Who should read this book - Overview - what is coming Introduction - September 2008: The Billion Dollar Day - January 2009: Why (most) humans make poor traders - The black box is simpler than you think - An open source revolution - An open source systematic framework PART ONE: THEORY The good, the bad and the ugly of systematic trading - Humans should be great traders - in theory -- The death of rational economic man -- Why we run losses and stop out profits - Introducing a systematic rule for trading -- Stick to the rules and don't meddle -- Overcoming instinct - why 'contra' instinctive behaviour works -- Why subjective 'systems' don't work - Commitment mechanisms - how do we stop ourselves 'meddling'? -- Automation - the use of dogs in finance and engineering - Systematic trading in financial institutions The commitment problem does not go away... -- . but there are benefits - The ideal systematic trading shop - Two more tricks to reduce meddling -- Abstraction -- Ignorance - Designing systems to discourage meddling - the three virtues -- Trust your system -- Understand the limits of your ignorance -- Sleep at night: Position size is as important than position sign - When is meddling acceptable? -- Unacceptable meddling -- Acceptable meddling - Irrationality in trading system development - the three sins -- Overfitting -- Overtrading -- Overbetting Systematic strategies - Why do strategies 'work'? -- Risk premia -- Frictions and barriers to entry -- Information less trading -- Returns to effort and cost -- Behavioural effects -- Pure alpha:skill - What makes a good strategy? -- Intuitive -- Well motivated -- As simple as possible -- Can be systematised - Categorising the strategy universe -- Static versus Dynamic -- Buying and selling insurance -- Technical vs fundamental -- Fast vs Slow -- Directional vs cross sectional -- Low versus high leverage -- Many positions vs few positions -- Crowd following vs contrarian PART TWO: THE TOOLBOX - Model selection, calibration and fitting -- The perils of overfitting -- Distinguishing dud models from good models -- Fitting and overfitting - Four rules for effective fitting -- Start with a small number of ideas, not with data -- Save real data for a rainy day; use artificial data -- Don't fit unless there is a gun to your head -- If you must fit to real data, be very, very careful - Portfolio allocation -- Anecdote: When smart people make stupid decisions -- The bad news: Portfolio optimisation is hard -- A simple fix: bootstrapping -- 'Handcrafting' the weights: The heuristic method -- Some problems -- The good news PART THREE: THE FRAMEWORK An 'open source' framework for systematic trading and investing - Why an open source framework? -- Parallels with open source software -- Flexibility -- Individual seperable components with well defined interface -- Underlying logic exposed a+' easily modified - The elements of the framework -- Instruments to trade -- One or more signals -- Forecasts - combinations of signals -- Scaled positions -- Portfolios of positions -- Total capital scaling - money management -- Risk measurement and control - Modifying and extending the 'open source' framework Instruments - the building blocks - Asset classes: Stocks, bonds, ETF's, futures, CFD's ... - The character of different instruments - Portfolios as instruments - Spreads - a special kind of portfolio instrument - Summary - key points Signals - looking under the hood - What is a signal? - What properties should signals have? -- A signal is a scaled quantity -- But what scale -- Why it makes sense to have a unit variance signal -- Are jumpy signals okay? -- Should we allow signals to be as large as possible? -- Three signals in detail -- Summary From signals to forecast - Combination -- Linear versus non linear -- Choosing the weights - we need portfolio optimisation -- The diversification multiplier - Mapping function -- Binary -- Linear -- Linear with cutoffs (recommended) -- Linear with flat spot - Summary - the default system does... Position scaling - The magic number - Position is signal over standard deviation - Expected volatility -- How do we measure expected volatility? -- Dangers of low volatility -- A rule for low volatility - Summary - the default system position scaling is... Instrument weights - more portfolio allocation - Linear weighting for portfolios - Portfolio optimisation amongst instruments -- Which grouping for the heuristic? -- Multiple dimensions -- Portfolios of spreads - The diversification multiplier, part two - Summary Total capital scaling: Risk appetite and money management - How much can you lose? - A brief primer on the Kelly Criteria - From Sharpe to Kelly - The total capital scaling rule -- Low risk target, high worst loss; or high risk target, low worst loss? -- Upside ratcheting and downside adjustment - Special cases: Interest paying, living off the proceeds and principal protection - Summary Risk measurement and risk control - Some risk management issues -- What is risk and how do we measure it? -- Risk that's hard to measure -- Two key flavours of system for risk management - Built in risk management -- Risk managing at a signal level -- Risk managing at an instrument level - System level risk management -- Maximum estimated risk -- Correlation risk - the perfect storm -- Jump risk redux - low volatility -- Combining them - the worst case scenario multiplier -- The clipping problem - Outside the system - the risk envelope -- The risk envelope exists to avoid meddling -- Measuring the envelope -- Applying the envelope - Buying an insurance against poor performance - Summary Tailoring - Speed of trading -- Calculating the damage from trading too quickly -- Decomposing and calculating the cost of trading -- Applying the brakes - how to slow down -- Costs and calibration -- Some subtleties - Trading with more or less capital -- Trading with more capital -- Trading with less capital PART FOUR: PRACTICE Example one: Systematic trading for discretionary traders - Why use a systematic framework with discretionary decisions? -- Instruments -- Signals -- Forecasts -- Position scaling -- A 'portfolio' of trades - Total capital scaling - Risk control - Worked portfolio example - Extensions Example two: Systematic asset allocation; a long only risk parity portfolio - A risk parity system - Instruments to trade - World's dullest signal and forecast - Position scaling - Portfolio construction - the difficult part -- Bootstrap method -- Heuristic method - Total capital scaling - Risk control - Worked portfolio example - Extensions Example three: Fully systematic futures trading system - A futures system - Instruments - Signals -- Momentum -- Carry - Combining signals to get forecasts -- Cost estimation -- Heuristic -- Bootstrapping - Position scaling - A portfolio of instruments -- Heuristic -- Bootstrap - Total capital scaling - the dangers of easy leverage - Risk measurement and control - Worked portfolio example - Extensions Appendices Appendix A: Resources Further reading Data sources Brokers and platforms Coding Appendix B: Formulas Backtesting - Accounting - Costs - Judging the results -- Sharpe ratio -- T- test -- Skew Fitting Iterative binary grid search Portfolio construction - Markowitz portfolio optimisation - Bootstrapped portfolio optimisation - By hand portfolio optimisation -- Means -- Costs - specific case of means - Linear portfolio weighting and calculating the diversification effect - Nearest portfolio Signals - Random entry stop loss - Flip flop stop loss - Basic moving average crossover - Exponetial moving average crossover - Raw carry signal for generic asset - Raw carry signal futures contracts - Smoothed carry signal From signal to forecast - Individual signal scaling - Linear signal combination and calculating the diversification effect - Forecast mapping functions - Linear with cap -- Binary -- Cutoff - Position scaling -- Volatility estimation - Minimum volatility rule - Final position calculation - Portfolios of instruments - Total capital scaling -- Establishing the initial scalar -- The auto ratchet down -- The manual ratchet up - Risk measurement and control -- Natural risk scalar -- Vol shock risk scalar - Correlation shock scalar - Total risk scalar - System performance envelope