Overfitting / curve fitting

Tuning a strategy's parameters to past data so tightly that it stops working going forward; the biggest trap of backtesting.

Signs: many parameters, results that collapse under small changes, perfect results in a single market or period. A strategy with few rules that works reasonably well across different markets and periods beats one that is perfect in a single place.

Safeguards: out-of-sample testing, walk-forward analysis, a parameter robustness sweep.