Validation answers one question: would this strategy have made money for reasons that will still hold, or did the history happen to suit it? We test strategies built by your team or by others, to the research standard we apply to our own work. If a strategy fails, the report says so, and says why.
What the work covers
- Validation procedures for manual, semi-automated and automated strategies
- Robustness testing on out-of-sample data, parameter neighbourhoods and cost assumptions
- Controls for data-mining bias when many variations have been tried, including the deflated Sharpe ratio and the probability of backtest overfitting
- Independent review of existing research, backtests and code
To discuss a project, write to our contact address.