Metrics6 min read

The Sharpe Ratio

The Sharpe ratio is the most widely cited performance metric in systematic trading. It is also widely misunderstood, misapplied, and gamed. Knowing what it measures — and what it doesn't — is foundational for interpreting any backtest result.

The formula

The Sharpe ratio measures risk-adjusted return: excess return per unit of volatility.

SR = (E[Rₚ] − Rƒ) / σₚ

Where E[Rₚ] is the expected portfolio return, Rƒ is the risk-free rate, and σₚ is the standard deviation of portfolio returns. In practice, daily returns are usually used with the daily risk-free rate approximated as zero for short horizons.

To annualise from daily returns:

Annual SR ≈ Daily SR × √252

The √252 multiplier is often misapplied. It assumes returns are independent and identically distributed, which they rarely are. Autocorrelated returns (common in trend-following strategies) require a different scaling factor.

What it measures

The Sharpe ratio captures the trade-off between return and volatility. A strategy that returns 15% per year with 10% volatility (SR ≈ 1.5) is preferable to one that returns 20% per year with 20% volatility (SR ≈ 1.0) if your goal is maximising return per unit of risk. This is why SR is useful for comparing strategies with different return profiles.

What it doesn't measure

Tail risk: The Sharpe ratio treats upside and downside volatility equally. A strategy that makes money steadily 95% of the time but has occasional catastrophic losses can show a high Sharpe ratio until the disaster occurs. Skewness and kurtosis — captured in the PSR and DSR calculations — are not reflected in the standard Sharpe.

Drawdown duration: A strategy can have a high Sharpe but a 2-year drawdown period. The Sharpe ratio gives no information about how long you'd have to wait through a losing stretch. The Calmar ratio (annual return / max drawdown) and maximum drawdown duration fill this gap.

Execution reality: A backtest Sharpe is computed on clean historical prices. Slippage, market impact, bid-ask spread, and position sizing constraints in live trading typically reduce the effective Sharpe by 20–50% for liquid instruments and more for less liquid ones.

As a rough benchmark: a Sharpe above 2 on 3+ years of live trading (not backtesting) is exceptional. Above 1.5 is strong. Between 0.75 and 1.5 is workable. Below 0.75 is marginal. Backtested Sharpe ratios are typically 2–3× higher than live Sharpe ratios for the same strategy, due to overfitting and execution shortfall.

Common abuses

Selecting the best in-sample Sharpe: If you optimise parameters to maximise Sharpe over historical data, the Sharpe you're showing is the maximum over many trials, not an unbiased estimate of future performance. The DSR corrects for this.

Short data periods: With fewer than 252 daily observations, the standard error of the Sharpe estimator is large enough that a “strong” result may be indistinguishable from noise.

Ignoring costs: Many retail backtests use zero or minimal transaction costs. A strategy trading 200 times per year with a 0.1% round-trip cost loses 20% per year to friction alone — visible in the post-cost Sharpe ratio but not in the gross.

Alternatives to consider

The Sortino ratio uses only downside deviation in the denominator, making it more sensitive to bad outcomes. The Calmar ratio anchors to maximum drawdown. The PSR and DSR provide statistical grounding. None of these replaces the Sharpe — they complement it. Kestrel Signalcomputes all of them on every result.