Out-of-sample evaluation of a trading rule's historical performance — strictly distinct from in-sample fitting. The four backtest killers: (i) look-ahead bias (using data not available at decision time, e.g. restated fundamentals or close prices stamped at the open), (ii) survivorship bias (delisted firms missing from the universe), (iii) data-mining bias / multiple-hypothesis testing (the deflated Sharpe ratio DSR corrects for the number of trials), and (iv) regime-dependence. The Bailey-López de Prado 'minimum backtest length' formula gives the number of independent trials needed before a reported Sharpe is statistically distinguishable from luck.