BacktestFolio · Glossary

Look-ahead Bias

In plain language

If you optimise a portfolio using 30 years of data and then 'test' the strategy on those same 30 years, you are using future information that did not exist at the time. It is like taking an exam having already seen the answers — results will always be excellent but not replicable going forward.

Technical definition

Methodological distortion in which a simulation uses information not available at the time of the simulated decision — such as ex-post corrected historical series or weights optimised with future data. Makes backtest results unrealistically positive.