Probabilistic Sharpe ratio (PSR)

The probabilistic Sharpe ratio (Bailey & López de Prado, 2012) converts a point estimate of the Sharpe ratio into a probability: given how long the track record is and how non-normal the returns are, what is the probability that the true Sharpe exceeds a chosen benchmark (often zero)? Skewed, fat-tailed returns and short samples widen the estimator's error, and PSR prices that in explicitly.

A strategy showing Sharpe 1.2 over 15 months of skewed returns can easily carry a PSR far below the certainty its point estimate suggests. As a rule of thumb, a PSR below ~95% against a zero benchmark means the sample cannot yet distinguish the strategy from noise.