Somebody hands you a strategy with a Sharpe of 1.8 and you're supposed to be impressed. Fair enough, that's the whole point of the number. It takes the return you earned above the risk-free rate, divides by the standard deviation of those returns, and gives you return per unit of volatility. A Sharpe of 1.0 means one point of excess return for every point of standard deviation. Higher is better, and the reason everyone uses it is that it lets you line up wildly different strategies on the same risk-adjusted footing.
William Sharpe published it in 1966 and it never left. Hedge fund decks, mutual fund fact sheets, pension evaluations, all of them lead with a Sharpe ratio. But the same simplicity that makes it travel so well also buries a few assumptions, and if you don't know what the number is quietly assuming, it will point you at the wrong strategy.
Standard deviation doesn't care which way you moved
The denominator treats a great month and a terrible month the same way. A strategy that occasionally rips off a huge positive return has a fat standard deviation, and that gets penalized exactly as hard as a strategy that occasionally craters. In practice this matters a lot. Think about a trend follower that grinds along with modest months and then catches one enormous move during a crisis. Its Sharpe looks mediocre because those big winners inflate the denominator. Meanwhile a short-vol strategy that collects small premiums like clockwork and blows up once every few years shows a beautiful Sharpe right up until the year it doesn't.
This isn't a corner case. It describes a real chunk of what trades out there. If you're comparing a convex strategy, one that loves big moves, against a concave one that hates them, the Sharpe ratio will quietly favor the concave strategy every single time until the concave strategy detonates.
The Sortino ratio fixes part of this by using downside deviation instead of total standard deviation, so it only penalizes the moves you actually care about. It's the better metric for anything with an asymmetric payoff. I tend to look at both side by side, because Sharpe and Sortino disagreeing is itself information about the shape of the returns.
The number depends on when you measured it
Sharpe ratios are more sensitive to the sample window than people expect. The same strategy can show a Sharpe of 2.0 across 2017 to 2019 and a Sharpe of 0.3 across 2017 to 2022, purely because the later window caught a regime the strategy hated. The number you're shown is always contingent on the period behind it, and two honest periods can tell you completely opposite stories.
Frequency matters too. Monthly and daily Sharpe ratios for the same strategy over the same span won't match. The annualized version is conventionally the per-period Sharpe times the square root of the number of periods per year, so a monthly Sharpe times the square root of 12. That scaling assumes returns are independent and identically distributed period to period, which they mostly aren't. Autocorrelation, which shows up constantly in trend-following and anything illiquid, makes the annualized Sharpe overstate or understate the real picture depending on which way the autocorrelation runs.
What you subtract as risk-free
Picking the risk-free rate sounds trivial and then quietly changes your answer, especially at the extremes of the rate cycle. Through the 2010s, when T-bill yields sat near zero, subtracting the risk-free rate barely touched the numerator, so you were basically Sharpe-ing raw returns. In a higher-rate world the same nominal returns produce a smaller excess and therefore a lower Sharpe. So when you compare Sharpe ratios across eras, remember the risk-free rate itself moved underneath them.
For anything international, the right risk-free rate is currency-specific. A Japanese equity strategy should be measured against JGB yields, not US Treasuries. Grab the wrong one and you can make a strategy look meaningfully better or worse than it really is.
Rough benchmarks so the number means something
A Sharpe is useless without a sense of scale. The S&P 500 has run somewhere around 0.4 to 0.5 over long stretches of 50 years or more, depending on how you slice it. A 60/40 stock and bond portfolio has historically landed around 0.5 to 0.7. Top-decile hedge funds report 1.0 to 2.0, though those figures lean hard on survivorship bias, backfill bias, and the smoothing you get from illiquid or infrequently priced positions.
- A sustained Sharpe above 2.0 over many years in liquid, tradeable stuff is genuinely rare.
- A backtest showing 3.0 or higher almost always means overfitting, data snooping, or a methodology bug hiding somewhere.
- Real frictions like transaction costs, slippage, and market impact reliably knock 30 to 50 percent or more off a backtested Sharpe once you trade it live.
That last one is worth sitting with. Half of a pretty backtest routinely evaporates on contact with a real order book, which is why I trust a modest Sharpe that survived live trading far more than a gaudy one that only ever existed in a simulation.
Using it without fooling yourself
Sharpe is at its best comparing strategies with similar return shapes over similar windows with the same risk-free rate. Two long-only equity strategies over the same five years, measured the same way, give you a real relative read. A long-only equity strategy's 10-year Sharpe against a market-neutral fixed income fund's 3-year Sharpe tells you almost nothing.
When you're sizing up a single strategy, put the Sharpe next to max drawdown, skew, and kurtosis. A Sharpe of 1.0 with an 8 percent worst drawdown is a totally different animal from a Sharpe of 1.0 that went through a 40 percent hole, even though the headline reads identical. Treat the ratio as the first question you ask, not the last, and you'll avoid the mistake most people make with it.