My trading system has a 73.5% win rate and loses money. Here is the diagnostic that found it.

Short summary

A 16-year FX backtest with 73.5% win rate loses money because small targets hit often but reward:risk is 0.25–0.4, making breakeven win rate 75–82%. The real failure was modelling costs post-hoc instead of at the trigger level, inflating win rate by 2–5 percentage points. A Spearman rank-correlation falsifier across 119 walk-forward folds returned median rho of -0.024, proving the parameter surface contained zero learnable signal.

• •73.5% win rate loses money because reward:risk ratio requires 75–82% breakeven
• •Post-hoc cost adjustment inflates win counts; trigger-level cost modelling changes which trades win
• •Spearman rank correlation across 119 folds showed zero learnable signal in the parameter surface

Generated with AI, which can make mistakes.