To backtest means applying a predefined strategy to historical data to simulate what would have happened under those rules. The result also depends on information available at the time, costs, orders, and constraints: it is more than running a signal over an old price chart.
How the term is used
“I backtested the strategy” should mean that the data, period, parameters, and execution model are stated. The verb alone does not certify the quality of the test or turn a simulation into a live performance record.
Technical distinction
A complete backtest looks backward and must prevent future information from entering decisions while modelling plausible costs and execution. A forward test instead observes a frozen version as new data arrive. A favourable historical result remains a hypothesis to validate, not a guaranteed forecast.
Sources
- Robert D. Arnott, Campbell R. Harvey, and Harry Markowitz, A Backtesting Protocol in the Era of Machine Learning — A recognized protocol covering data, model selection, and limits of inference from backtests.
- European Commission, Commission Delegated Regulation (EU) 2017/589, RTS 6 — Official testing, documentation, and separated-environment requirements within its regulatory scope.