Backtesting

Backtesting runs a defined set of rules over past price data to see what results they would have produced. Done carefully it reveals a strategy's expectancy, its worst losing run and its behaviour across different market regimes.

Its failure modes are well known: hindsight bias from seeing future bars, curve fitting parameters to one specific history, and omitting realistic spread, commission and slippage. A test that ignores costs and reserves no out-of-sample data manufactures confidence rather than evidence.