Sharpe, Sortino, and what a ratio can and cannot tell you

Risk-adjusted return ratios compress a whole track record into one number. That is their use and their limitation.

Sharpe ratio

Excess return over the risk-free rate divided by the standard deviation of returns. It penalises upside volatility as much as downside, which is defensible for a symmetric strategy and misleading for an asymmetric one.

Sortino ratio

The same numerator over downside deviation only. It is more forgiving of strategies with occasional large gains. It is also easier to inflate with a short sample that happens to contain no large losses.

Sample size

A Sharpe ratio measured over three years has a standard error close to 0.6. A reported Sharpe of 1.0 over that horizon is consistent with a true Sharpe anywhere from roughly 0.4 to 1.6. Treat any ratio without a horizon and a confidence interval as a headline, not a finding.