Backtesting: reviewing trades you never took

Journaling & Review

Replaying the past to test a rule

Backtesting means scrolling back through historical charts, finding every place your setup occurred, and recording what would have happened — before risking money on it. It's the only way to get a hundred samples of a setup without spending a year collecting them. The rule has to be written down first and it has to be specific enough that someone else could apply it and get your answers: 'buy pullbacks in an uptrend' is not testable, 'buy the first close above the 20 EMA after price has made a higher high and a higher low' is.

The two ways everyone cheats

Hindsight bias: you already know what happened next, so you 'find' entries that you'd never have taken live. Cure — cover the right side of the screen and step forward one candle at a time. Curve fitting: you keep adding conditions until the past looks perfect, and end up with a rule that describes one specific year rather than a market. Cure — a rule with four conditions that wins 55% beats one with eleven conditions that wins 80% on the sample you built it from.

What a backtest can and cannot tell you

It can tell you a setup's win rate, average R, worst losing streak, and whether the expectancy is positive at all. It cannot tell you whether you'll actually take the trade at 9:47 on a red morning, and it cannot include the spread and fees unless you deliberately subtract them — which is where most impressive backtests quietly die. Wick's arcade is a backtest with the cheating removed: the future is hidden, the costs are charged, and the result goes into the same Edge Score as everything else.