Deflated Sharpe ratio (DSR)

If you test enough variants, the best backtest will look brilliant by construction: the expected maximum Sharpe among N random trials grows with N even when every trial has zero true skill. The deflated Sharpe ratio (Bailey & López de Prado, 2014) raises the benchmark accordingly: it computes the Sharpe you would expect from pure selection among your trials, and asks whether the winner clears it.

The practical consequence is uncomfortable and healthy: the more configurations you searched — parameter grids, filters, weekday exclusions — the higher the bar your final pick must clear. A strategy that only looks good because it won a large tournament of siblings is deflated back to what it is.