The open research series: building a systematic strategy in public

An "open-source hedge fund" is less a fund than a way of working: publishing the reasoning, code and test results behind a systematic strategy so that others can check, reproduce and improve them. This page collects our articles on that idea in the order we suggest reading them. They are educational; no fund is offered here.

1. The idea

  1. What is an open-source hedge fund?Crowdsourced and open research models, what they gained and where they stopped.
  2. Running a research project in the openVersion control, reproducibility, licences and review for a public research programme.
  3. Creating a strategy, step by stepFrom hypothesis to paper trading, with the checks that belong at each step.

2. The worked example: momentum

  1. Momentum strategiesCross-sectional and time-series momentum and the evidence for them.
  2. How many stocks, and how to weight themDiversification, concentration and weighting schemes.
  3. Momentum crashesWhen the strategy fails, and why it fails so abruptly.
  4. Market timing models for a momentum strategyTrend filters and regime signals as a defence against crashes.
  5. Momentum on dual momentum portfoliosApplying the momentum rule one level up, to whole portfolios.
  6. The Fama-French five-factor modelChecking whether a return is new or a known factor in disguise.

3. Risk and combination

  1. Choosing your riskVolatility targets, drawdown limits and position sizing.
  2. An introduction to value at riskHistorical, parametric and Monte Carlo VaR, and expected shortfall.
  3. Hidden Markov models for market regimesEstimating unobservable market states from returns.
  4. Combining diversified alphaWhy several weak, uncorrelated strategies can beat one strong one.

4. Filters from machine learning

  1. A laboratory for machine learning in financeBars, labels and validation suited to market data.
  2. Meta-labelling: a toy exampleA secondary model that decides whether to act on a signal.
  3. Does meta-labelling add to signal efficacy?How to test the claim without fooling yourself.

How to use the series

Each article stands on its own, with its references at the foot of the page. Where an article includes numbers to illustrate a method, it states whether they are hypothetical or taken from a cited study. The series does not publish live signals, model portfolios or performance claims, and nothing in it is a recommendation to invest. Suggestions of sources and corrections are welcome; see how to contribute.