The portion of historical data set aside not used during strategy development, but later used to validate the strategy’s performance. After optimizing a strategy on in-sample data, traders test it on out-of-sample data – which is effectively unseen data – to see if it still performs well. This simulates how the strategy might behave on future data. Out-of-sample backtesting is critical because it shows whether a strategy holds up on “unknown” data. For example, if 2015–2020 was in-sample, then 2021–2022 could be out-of-sample to evaluate if the strategy’s profitability and drawdowns remain acceptable.

Back to Robots Glossary Index