Sharpe ratio

The Sharpe ratio measures return per unit of risk: take a strategy's average return, subtract the risk-free rate, divide by the volatility of those returns. A strategy earning 20% a year with wild swings can score worse than one earning 12% smoothly. The metric exists specifically to price that volatility.

Raw return tells you nothing about how you got there, and win rate tells you even less. A system can win 90% of its trades and still end badly, because the losses it does take wipe out years of gains — the failure mode we call tail-concentrated risk. Sharpe forces that volatility into the number, which is why it's the closest thing backtesting has to a single quality score. Closest thing. Not a clean one.

The ratio assumes returns cluster around a normal, bell-curve distribution. Most trading strategies don't. A grid or DCA-style system can post months of tiny, steady gains and then one gap-down loss that erases them all. Right up until that event the Sharpe reads smooth, because the drawdown that eventually shows up hadn't happened yet inside the sample being measured.

Serial correlation makes it worse. Strategies that hold positions across many bars, or trade instruments that don't move independently day to day, can look steadier than they really are. That understates the true volatility and inflates the annualized Sharpe well past what live trading would ever deliver.

Then there's how many strategies you tried to get the one you're looking at. Test 500 variations of an idea and rank them by Sharpe, and the top result is partly skill and partly the luck of trying 500 times. A single backtest's Sharpe, reported alone with no accounting for that search, says almost nothing about whether the edge survives outside the sample it was measured on.

We don't take a headline Sharpe at face value. Every candidate that clears an initial test gets checked out-of-sample, and every survivor goes through an adversarial audit for look-ahead bias before we call anything real. A flattering Sharpe next to a weak profit factor or a tail-concentrated return stream tells us the number was dressed up. The ratio is useful. It's also one of the easiest statistics in trading to flatter without changing a single trade.