Guides / Sharpe Ratio vs. Sortino Ratio: Which One Should You Trust?

Risk-adjusted return · two formulas

Sharpe ratio vs. Sortino ratio, what the difference actually means.

Both claim to measure risk-adjusted return, but they're built on a different idea of what counts as risk. Here's the actual formula difference, and which ratio a strategy with an occasional big winner can hide behind.

Cephic Team Updated Nov 2, 2026 8 min read

Quick answer: Sharpe divides excess return by the standard deviation of every return in the sample, so up days and down days both count as "risk." Sortino divides excess return by the standard deviation of only the returns that fall below a target (usually 0% or the risk-free rate), so upside swings never touch the denominator. A strategy with a few outsized winning trades and a tame downside will show a much higher Sortino than Sharpe, which is exactly why the two ratios can tell conflicting stories about the same track record.

THE CORE FORMULA DIFFERENCE

What each ratio actually penalizes

Sharpe ratio = (mean return − risk-free rate) / standard deviation of all returns. Every return in the sample (the quiet days, the losing days, and the one day the strategy made ten times its usual amount) goes into that standard deviation. A strategy doesn't need to lose money to get a worse Sharpe ratio; it just needs returns that vary, in either direction.

Sortino ratio = (mean return − target return) / downside deviation. The target is usually 0% or the risk-free rate, and "downside deviation" only measures the spread of returns that fall below it. Returns above the target (including a single huge winning trade) still push the numerator (the mean) up, but they don't appear in the denominator at all, so they can't drag the ratio down the way they would under Sharpe.

That's the entire difference. Same numerator logic (excess return over some baseline), but a different denominator: all volatility, versus only the volatility that actually hurt.

WHY THIS MATTERS WHEN COMPARING STRATEGIES

Why the choice changes which strategy looks better

Picture a strategy with a long right tail: most trades are small and unremarkable, but once every month or two it catches a real trend and books a trade several times the size of a typical winner. Its downside stays tightly controlled: small, consistent losses, nothing dramatic. On Sharpe, that one outsized winner inflates the standard deviation of the whole return series, and the ratio gets penalized for it exactly as if it had been an outsized loss. On Sortino, that same trade is invisible to the denominator. The ratio only sees the small, controlled losses, and comes out looking much better.

This is a real, well-documented critique of Sharpe, not a matter of opinion, and it genuinely cuts both ways. Sharpe's defenders have a point: volatility of any kind, including sudden unexplained upside, can be a sign of an unstable or poorly understood process. A strategy that occasionally produces a trade five times its normal size might also be one regime shift away from a loss five times its normal size, since the same mechanism that produces the win could just as easily produce the opposite. Punishing that uncertainty isn't unreasonable.

Sortino's defenders counter that this reasoning doesn't match how traders actually think about risk. Nobody closes out a strategy because it made too much money too quickly. Treating upside variance as equivalent to downside variance conflates two things traders care about very differently, and a ratio that's supposed to measure risk-adjusted return shouldn't punish the return part of that equation. Which argument you find more convincing depends on whether you trust the upside to keep showing up, and that's not something either ratio can tell you by itself.

A WORKED EXAMPLE

Two strategies, same average return, different ratios

Take two hypothetical strategies with the exact same average daily return over a simple five-day stretch, so the only thing that differs is the shape of the returns, not the underlying edge. The risk-free rate and Sortino's target are both set to 0% for simplicity, which is a common simplification when working the formulas by hand.

Strategy A ("Lumpy"): daily returns of −1%, −1%, 1%, 1%, 10%. Mean return: (−1 − 1 + 1 + 1 + 10) / 5 = 2%.

Standard deviation, against that 2% mean: deviations are −3, −3, −1, −1, 8. Squared: 9, 9, 1, 1, 64, sum 84. Variance = 84 / 5 = 16.8. Standard deviation = √16.8 ≈ 4.10. Sharpe A = 2 / 4.10 ≈ 0.49.

Downside deviation only counts the two −1% days; every other day contributes zero to this calculation. Squared downside deviations: 1 and 1, sum 2. Downside variance = 2 / 5 = 0.4. Downside deviation = √0.4 ≈ 0.63. Sortino A = 2 / 0.63 ≈ 3.16.

Strategy B ("Smooth"): daily returns of −1%, 1%, 2%, 3%, 5%. Same mean: (−1 + 1 + 2 + 3 + 5) / 5 = 2%, and the same single −1% losing day as Strategy A. The only difference is how the other four days are spread out.

Standard deviation: deviations from the mean are −3, −1, 0, 1, 3. Squared: 9, 1, 0, 1, 9, sum 20. Variance = 20 / 5 = 4. Standard deviation = 2.00. Sharpe B = 2 / 2.00 = 1.00.

Downside deviation: only the −1% day counts, squared deviation = 1. Downside variance = 1 / 5 = 0.2. Downside deviation = √0.2 ≈ 0.45. Sortino B = 2 / 0.45 ≈ 4.47.

Line the four numbers up. On Sharpe, B beats A by more than double (1.00 versus 0.49) because A's one big winning day reads as instability. On Sortino, B still wins, but the gap closes to under 1.5x (4.47 versus 3.16) because both strategies share the identical downside, and Sortino simply doesn't care that A's upside was lumpier. Same two return streams, same risk-free assumption. How big A's apparent disadvantage looks depends entirely on which ratio is on the page.

WHERE THIS GETS MISUSED

Where this gets misused

WHAT TO ACTUALLY DO

What to actually do

Look at both numbers, not just the one that's quoted. If they diverge a lot, that divergence is itself informative: it tells you the strategy's volatility leans toward the upside, which is worth understanding before you size into it, not something to wave away because the more flattering ratio exists. Know what each one is actually penalizing before you compare two strategies by their ratios at all.

Neither ratio should stand in for actually looking at the equity curve and the drawdowns. A single number compresses a multi-year return history into one figure and necessarily throws information away: how long the worst drawdown lasted, whether the edge depends on a handful of outlier trades, whether the shape of returns has been drifting. Pull up the curve. Check the drawdown depth and duration separately. Treat Sharpe and Sortino as two inputs into that picture, not a replacement for it.

Allocate grades each strategy by what actually breaks it (profit concentration, how long drawdowns run, and whether the edge still survives with its biggest winners removed) instead of collapsing everything into a single ratio that's easy to cherry-pick.

Open Allocate