Sign up
ContentsHideShow

Quant glossary · Risk & performance

Probabilistic Sharpe ratio

The probabilistic Sharpe ratio is an approximate frequentist confidence score for a population Sharpe exceeding the benchmark.

Definition

The probabilistic Sharpe ratio is an approximate frequentist confidence score for a population Sharpe exceeding the benchmark, obtained from the estimator's sampling distribution. Define gamma3=m3/m2^(3/2) and gamma4=m4/m2^2 from central moments m_k=n^-1 sum_i(X_i-X-bar)^k, where X denotes excess return and gamma4 is raw kurtosis. The displayed formula is an asymptotic approximation for iid returns with positive variance, finite fourth moment and n>1; dependence needs an adjusted standard error. It is not a multiple-testing correction by itself.

How it is calculated

PSR=Φ((S^−S0)n−11−γ3S^+γ4−14S^2)PSR=\Phi\left(\frac{(\hat S-S_0)\sqrt{n-1}}{\sqrt{1-\gamma_3\hat S+\frac{\gamma_4-1}{4}\hat S^2}}\right)

Probabilistic Sharpe ratio = Phi of (observed Sharpe minus benchmark Sharpe) times the square root of (n minus 1), divided by the square root of 1 minus skewness times observed Sharpe plus (kurtosis minus 1) times observed Sharpe squared divided by 4.

Symbol Meaning Unit
PSR probabilistic Sharpe ratio probability
Phi standard normal CDF probability
S observed Sharpe ratio dimensionless
S^* benchmark Sharpe ratio dimensionless
n number of return observations count
gamma gamma3 raw skewness and gamma4 raw kurtosis from central moments dimensionless

Interpretation

PSR turns a noisy Sharpe comparison into a probability-like score: how much evidence does the sample provide against a chosen benchmark? Skewness and kurtosis change the uncertainty of that comparison. It answers a different question from correcting for a search over many trials.

Assumptions and conventions

  • The displayed finite-sample convention uses gamma3=m3/m2^(3/2) and gamma4=m4/m2^2, with m_k=n^-1 sum_i(X_i-X-bar)^k and X the excess return.

  • Returns are iid with positive variance, finite fourth moment, and n>1.

  • The formula is an asymptotic normal approximation; dependence needs an adjusted standard error.

PSR is a probability; all Sharpe and moment terms are dimensionless.

Limitations and common misreadings

  • PSR is not the probability that a strategy will make money.

  • The benchmark Sharpe must be stated; zero is not automatic.

References

  1. David Bailey, Marcos López de Prado. (2012). The Sharpe ratio efficient frontier. The Journal of Risk. Discussion introducing the probabilistic Sharpe ratio and its benchmark-comparison formula. doi:10.21314/jor.2012.255. Source

Probabilistic Sharpe ratio FAQ

What is Probabilistic Sharpe ratio?

The probabilistic Sharpe ratio is an approximate frequentist confidence score for a population Sharpe exceeding the benchmark.

How is Probabilistic Sharpe ratio calculated?

Probabilistic Sharpe ratio = Phi of (observed Sharpe minus benchmark Sharpe) times the square root of (n minus 1), divided by the square root of 1 minus skewness times observed Sharpe plus (kurtosis minus 1) times observed Sharpe squared divided by 4.

What does Probabilistic Sharpe ratio not tell you or assume?

PSR is not the probability that a strategy will make money.

Browse all glossary termsRisk and performance termsQuant research library

Definitions are educational and descriptive, not investment advice. Worked examples use labelled sample data.