"What could be wrong with this result?" not "How good is this result?" - "How would this fail?" not "How much would this make?" - "Is this edge robust across different time periods, markets, and parameter choices?" not "What parameters maximize the backtest return?"