How to test an investment strategy
A good story deserves a hard test.
A strategy needs a stated hypothesis, realistic costs and evaluation on data it has not seen. A strong historical chart is only the beginning.
Define success before you test
State the forecast target, horizon and decision rule. Distinguish predicting direction from earning a return after costs. Select comparison baselines before inspecting results and preserve the timestamp at which each input was actually available.
Keep a period genuinely unseen
Trying many features and choosing the best historical result can overfit noise. Use chronological training and validation windows, then evaluate once on a held-out test period. Overlapping target windows need special care to prevent information leakage.
Measure the uncomfortable parts
Report drawdowns, turnover, slippage, sample size and performance in different regimes. For probabilities, check calibration and Brier score against a base-rate forecast. Paper results and a 60–90-session shadow run are evidence to review, not a guarantee of future performance.
Try the idea
With no gross gains, repeated trading costs reduce the starting $100. A strategy test must include spreads, fees, slippage and taxes where applicable.
View values
Which result is the strongest reason to investigate further?
Choose an answer. Completion is saved only in this browser.