Out-of-sample test — Trading Glossary

Testing a strategy on data that was never used to design or adjust it — the only kind of backtest result that says anything about the future.

Split the history before starting: one part to design and compare changes, one part kept unseen and run exactly once at the end. A result on the design data partly measures how well the rules were fitted to it.

Also called: out-of-sample, unseen data.

Taught in

We Backtested It: The VP StrategyStrategy Playbook

The volume profile rules from the Reel, coded and tested on 21 months of US30 — why they lost, what we changed, and how we tested without fooling ourselves.

Also covered in that lesson

Curve fittingAdjusting a strategy's rules until its backtest looks good — which describes the past data rather than finding anything that will repeat.

Related terms

Curve fitting Sample size