Knowledge Centre · Start here
Introduction to the series
The field notes are a technical curriculum in systematic trading, from the first test of an idea to a strategy running with real money. Claims rest on public data and primary research, and figures come with their code.
Audience
Who the notes are for
We write for engineers, analysts and students who plan to research and run systematic strategies themselves. We assume you are comfortable with Python and pandas, with basic statistics and probability, and with the material of a first course in finance: returns, portfolios, leverage and the Sharpe ratio.
The notes are not an introduction to investing and contain no trade ideas. If Python or basic statistics are new to you, learn those first. Everything here is general, and a technique that suits one portfolio can be wrong for another.
Tracks
Five tracks, one reading path
Each track answers one question a working trader faces. Solid boxes are published; dashed boxes are in preparation.
Foundations
Which ideas deserve to be traded, and how should they be tested?
Research infrastructure
What runs a backtest, and how do you keep thousands of them reproducible?
- QuantConnect and LEAN
- Running backtests at scaleIn preparation
Portfolio construction and risk
How large should each position and each strategy be, and what limits the downside?
- Volatility targeting
- Building strategies from componentsIn preparation
- Combining strategies into one bookIn preparation
Production operations
Once real money is at stake, how do you know the system is doing what the research said it would?
- Monitoring a live bookIn preparation
- Alerting that people act onIn preparation
- Live versus backtest reconciliationIn preparation
Reading path
The reading path
Later notes assume the earlier ones unless their prerequisites say otherwise.
- FoundationsMarket anomaliesWhich return sources have held up, and why do most published ones fade?
- FoundationsIndicators and the 200-day moving averageHow do you turn an idea into a rule, and test it without look-ahead or hidden costs?
- Research infrastructureQuantConnect and LEANWhat engine runs one algorithm in backtest and live, and what must you check before trusting it?
- Research infrastructure · in preparationRunning backtests at scaleHow do you run thousands of backtests and still trust the results?
- Portfolio construction and riskVolatility targetingHow large should a position be?
- Portfolio construction and risk · in preparationBuilding strategies from componentsHow do you build a strategy from parts that can be tested and replaced separately?
- Portfolio construction and risk · in preparationCombining strategies into one bookHow much capital should each strategy get, and what happens when strategies trade the same instruments?
- ExecutionTWAP, slicing and jitterWhat does a trade really cost, and how do you measure it?
- Production operations · in preparationMonitoring a live bookWhat should be checked every trading day once real money is at stake?
- Production operations · in preparationAlerting that people act onWhich conditions deserve to interrupt someone, and how do you keep alerts trusted?
- Production operations · in preparationLive versus backtest reconciliationWhy do live results differ from the simulation, and when does the difference matter?
Shortcuts
- New to systematic research. Start with Market anomalies and follow the path.
- Judging a backtest someone else ran. Read the section on data snooping in Market anomalies, then the six checks in Indicators and the 200-day moving average.
- Choosing or auditing a backtesting engine. Start with QuantConnect and LEAN.
- You have a signal and need to size it. Start with Volatility targeting.
- Already trading, and want better execution. Start with TWAP, slicing and jitter.
Inside a note
What to look for
- Abstract and key takeaways
- The first screen of every note states the argument and its conclusions. If the takeaways are familiar, skip to the sections that are not.
- Before you start
- The background each note assumes, and the notes it builds on.
- Notation
- Symbols mean the same thing in every note; the table below is the reference.
- Code listings
- Python with pandas and NumPy, labelled with the file each listing comes from. Where a listing produced a number, it ran on the public data or the simulator named beside it, so you can reproduce the figure.
- Interactive labs
- Each lab's caption suggests experiments. Change one setting at a time, and predict the result before you look.
- Evidence and caveats
- Results come with their limits: the sample, the sensitivity to parameters, and the conditions under which a technique fails.
- References
- Primary sources, cited where they are used and listed at the end with a DOI or a link.
Standards
How the notes are written
- Claims about markets rest on public data or primary research, and the source is named where the claim is made.
- Signals use only data available at the time, and costs are modelled rather than assumed away.
- Failures, decay and weak periods are reported alongside what worked.
- Every figure built from data can be reproduced from the code shown with it.
- The notes describe published methods, not the rules or parameters of our strategies.
Notation used throughout
| rₜ | simple return over period t (daily unless stated) |
| Pₜ | price or total-return index at the close of t |
| wₜ | exposure held over period t, decided at the close of t − 1 |
| σ, σ̂ₜ | annualised volatility; its forecast made with data up to t − 1 |
| SR | Sharpe ratio: mean excess return over its standard deviation, annualised |
| T | length of a sample in years; N the number of trials, slices or observations |
| bp | basis point, 0.01% |
The series
How it grows
Notes are cited by title, and their addresses do not change. New notes take their place on the reading path as they are published.