What Is the Sharpe Ratio, and What Does It Actually Measure?
The Sharpe ratio is a track record's average return divided by the standard deviation of its returns — a single number for how much return a system produced per unit of bounce it took to get there. Two systems can post the identical average return and land on very different Sharpe ratios, because the ratio isn't scoring the return itself, it's scoring how smooth or jagged the ride to that return was.
That's the answer to the search query. The rest of this page covers how the ratio is actually calculated, why a smoother record scores higher even at the same average return, the specific blind spot that lets a dangerous strategy post an impressive-looking number, and what the ratio doesn't tell you on its own.
How is the Sharpe ratio calculated?
Take the average return over some period, subtract a risk-free rate (what the money could have earned doing nothing), and divide the result by the standard deviation of those same-period returns.
- Collect a return for each period (daily, monthly — whatever the underlying data uses), not just the final number.
- Average those returns.
- Subtract the risk-free rate for the same period, so the ratio measures return in excess of parking the money in cash.
- Divide by the standard deviation of the period returns.
- Annualize if needed by multiplying by the square root of the number of periods in a year.
The denominator is doing the real work here. A system that earns the same average return with less period-to-period bounce produces a bigger ratio than one that earns it with more bounce — the formula rewards consistency, not just size of return.
Why can two records with the same return have different Sharpe ratios?
Because the ratio divides by bounce, and two paths to the same destination can have very different amounts of it. A record that grinds out a small, steady gain most periods and rarely has a bad one produces a low standard deviation, which makes the ratio's denominator small and the ratio itself large. A record that earns the identical average return through a mix of big up periods and big down periods produces a high standard deviation, which makes the same average return divide down into a much smaller ratio.
This is the entire point of the metric — it's built to tell those two records apart when a bare average-return figure can't. See win rate vs. expectancy for the same idea applied one level earlier: a number that only reports the average, on its own, can describe two very different real experiences.
Where does the Sharpe ratio's blind spot actually bite?
Standard deviation treats a surprise gain and a surprise loss as the same kind of "bounce" — it has no idea which direction the deviation ran. A strategy built to collect a small, steady income most of the time while carrying a rare, severe loss (selling far out-of-the-money options is the classic shape) can produce years of smooth, low-bounce returns and therefore a high Sharpe ratio, right up until the rare loss finally shows up and erases a large chunk of the account in one stretch. The ratio measured every quiet period correctly; it simply has no mechanism for seeing a loss that hasn't happened yet.
| What the record looked like | Sharpe ratio read | What was actually true |
|---|---|---|
| Small steady gains, the big loss still unrealized | High | A rare severe loss was sized into the strategy and simply hadn't landed |
| Small steady gains, then one severe loss | Drops sharply, after the fact | The risk was there the whole time; only the backward-looking number changed |
Reading maximum drawdown alongside the Sharpe ratio catches part of this — see what maximum drawdown is and why it matters more than win rate — because a drawdown figure is at least asking "how bad did it get," even though neither number can see a loss that's still in the future.
What else does the Sharpe ratio not tell you?
- It doesn't distinguish upside bounce from downside bounce. A system with occasional sharp gains and otherwise flat returns gets penalized by the same standard-deviation math as one with occasional sharp losses — both read as "volatile," even though only one of them is the kind of surprise anyone actually worries about.
- It assumes returns are roughly symmetric and well-behaved. A return series with a long, fat left tail (frequent small gains, rare huge losses) violates the assumption the ratio's statistics are built on, which is exactly the shape of strategy the blind spot above describes.
- Sample size still applies. A short record hasn't lived through enough conditions to show its real volatility yet — see how many calls it takes before a track record means anything for the same caution applied to a win rate instead of a ratio.
- A backtested Sharpe ratio inherits every limitation the backtest has. A rule set searched until it happened to avoid the market's roughest stretches will report a smoother, higher-Sharpe history than the strategy's real forward risk — see what curve-fitting is and how it fakes a backtested track record.
- It says nothing about whether the record is genuine. A cherry-picked window, or an account that only shows its best stretch, can produce an inflated Sharpe ratio the same way it inflates any other summary statistic — see how to spot a cherry-picked track record.
Questions traders ask
Is a higher Sharpe ratio always better?
Not automatically. It's better at the specific thing the ratio measures — more return per unit of period-to-period bounce — but it can't see a loss that hasn't happened yet, and it penalizes upside bounce the same way it penalizes downside bounce. A strategy with a quietly-building tail risk can carry a high Sharpe ratio right up until the tail risk arrives.
What counts as a "good" Sharpe ratio?
There's no single universal cutoff, and a number quoted without its time period, asset class, and whether it's from a live account or a backtest doesn't mean much on its own. The more useful comparison is the same strategy's Sharpe ratio across different stretches of time, and against its own maximum drawdown and sample size, rather than against an arbitrary round number.
Is the Sharpe ratio the same as the Sortino ratio?
No, though they're built the same way. The Sortino ratio uses the same average-return-over-risk shape but only counts downside bounce in its denominator, which is the direct fix for the blind spot described above — a strategy with occasional large gains and otherwise steady returns scores better on Sortino than on Sharpe, since the gains stop being counted as a penalty.
Can a backtest's Sharpe ratio be trusted more than its win rate?
No more than any other single statistic pulled from the same backtest. Whatever made a win rate or an expectancy figure unreliable in a given backtest — a short sample, curve-fitting, missing real trading costs — makes the Sharpe ratio pulled from that same backtest just as unreliable, since all three numbers are computed from the identical set of returns.
Where this fits the bigger picture: a Sharpe ratio, a maximum drawdown, a sample size, and clean (not cherry-picked or curve-fit) data are the legs a track record needs before any single number from it means much. How the method behind this site's own graded calls holds up against that bar is on the methodology page.
Keep reading
Get them free