Skip to content
Backtesting & statistics

Deflated Sharpe Ratio (DSR)

Education only · our voice · free public data

Definition

The probability that a strategy's observed Sharpe is genuinely greater than zero AFTER correcting for the number of configurations tried, the length of the track record, and the skew/kurtosis of the returns. Reported as a probability in [0,1]; >0.95 is the usual keep threshold.

How to read it

When you try N strategy variants, the best one's Sharpe is inflated purely by selection, even if none has real edge. DSR estimates the 'benchmark' Sharpe you'd expect from the luckiest of N random trials and asks whether the observed Sharpe clears it with statistical confidence. A high raw Sharpe with a low DSR is the signature of data snooping: the strategy looks good only because you looked many times.

How practitioners use it

Used as context among multiple indicators — never as a standalone signal to act.

Less common professional uses

The expected maximum Sharpe of N independent trials grows like sqrt(variance of trial Sharpes) x [(1-g)*Z^-1(1-1/N) + g*Z^-1(1-1/(N*e))], where g is the Euler-Mascheroni constant and Z^-1 is the inverse normal CDF; this is the deflation benchmark SR0 subtracted before the significance test. DSR folds negative skew and excess kurtosis into the Sharpe's own standard error, so fat-tailed short-vol strategies are penalized twice: once via a higher SR0 hurdle and once via a wider error band. Effective-trials, not raw-trials, is what matters: correlated variants (e.g. RSI(13) vs RSI(14)) count as fewer than one independent trial each; estimate the effective N via the correlation structure of the trial returns, otherwise DSR over-penalizes. DSR is the single-strategy dual of the multiple-testing framework; for a whole research program prefer controlling the family-wise error or false-discovery rate across all backtests.

Sources & provenance

Bailey & Lopez de Prado (2014), 'The Deflated Sharpe Ratio', Journal of Portfolio Management; Lopez de Prado (2018), 'Advances in Financial Machine Learning', Ch. 8

This page is educational content published by Pachira Aquatica Global LLC. It is not investment advice and not a recommendation.

← All indicators