HomeTools › Probabilistic Sharpe Ratio calculator

Probabilistic Sharpe Ratio calculator

This calculator returns the Probabilistic Sharpe Ratio: the probability that a strategy's true Sharpe ratio exceeds a chosen benchmark, given the observed Sharpe, the track-record length, and the skewness and kurtosis of the returns. It runs entirely in your browser — nothing is uploaded or stored.

What it answers

A Sharpe ratio is an estimate from a sample. PSR asks the question that actually matters to an allocator: given this much data and this return shape, how confident can I be that the true Sharpe beats my benchmark? A PSR of 0.95 is the conventional bar — below it, the record does not support the claim at 95% confidence.

Calculator

Probabilistic Sharpe Ratio

Frequency consistency matters. SR̂, SR* and n must describe the same sampling frequency. If you enter an annualised Sharpe, n is still the number of observations (24 monthly returns = 24, not 2). Mixing them is the most common error with this formula.

Formula

( SR_hat - SR* ) * sqrt( n - 1 ) PSR(SR*) = Z [ ---------------------------------------------------- ] sqrt( 1 - g1*SR_hat + ((g2 - 1)/4)*SR_hat^2 ) Z standard normal CDF g1 skewness n number of observations g2 kurtosis (normal = 3)

Negative skew makes -g1*SR_hat positive, enlarging the denominator and lowering PSR. Excess kurtosis does the same. Both encode the fact that fat-tailed, negatively skewed returns make a Sharpe estimate less trustworthy — which is precisely the profile of strategies that look best on headline numbers.

Example calculation

The values the calculator loads by default, so you can verify it against an independent implementation before trusting it on your own numbers:

InputValue
Observed Sharpe SR̂1.50
Benchmark SR*0.00
Observations n24
Skewness γ₁−1.20
Kurtosis γ₂7.00
den = sqrt( 1 - (-1.20)(1.50) + ((7.00 - 1)/4)(1.50)^2 ) = sqrt( 1 + 1.80 + 1.5 * 2.25 ) = sqrt( 6.175 ) = 2.4850 z = (1.50 - 0.00) * sqrt(24 - 1) / 2.4850 = 1.50 * 4.7958 / 2.4850 = 2.8955 PSR = Z(2.8955) = 0.9981 -> 99.81%

Change the benchmark to 1.00 and the same track record gives z = 0.965 and PSR = 0.8327, which fails the 0.95 bar. Same strategy, different question.

Assumptions

Related