A backtest is a historical simulation, not proof of skill
A credible backtest recreates what could have been known and executed at the decision time. A model that uses revised stats, end-of-season rankings, final injury reports or unavailable odds to predict an earlier match will often look far stronger than it really is. This is look-ahead leakage. Other problems include selection bias, repeated tuning to the test set, survivorship bias and unrealistic execution assumptions.
