Sortino Ratio: Formula, Downside Deviation, and When to Use It Over Sharpe

Direct Answer

The Sortino ratio is (Rp − Rf) ÷ σd, where σd is downside deviation — the standard deviation calculated only from returns that fell below a chosen minimum acceptable return (MAR). Because it drops every period of upside volatility from the denominator, it is usually higher than the Sharpe ratio calculated on the same returns, and it is most useful for strategies whose return distributions are not symmetric.

The goal is not to memorize the formula in isolation. It is to understand what downside deviation actually measures, why the choice of MAR changes the number, and when reaching for Sortino instead of Sharpe answers the question that actually matters for a given strategy.

Key Takeaways

• Sortino divides excess return by downside deviation, not total standard deviation.
• Downside deviation counts only returns below the minimum acceptable return (MAR); upside volatility is excluded entirely.
• The MAR is a choice — 0%, the risk-free rate, or a target return — and it must be stated whenever the ratio is quoted.
• Sortino is typically higher than Sharpe on the same data, since its denominator is built from a smaller, downside-only slice of the returns.
• Sortino fits asymmetric strategies (options selling, trend-following, crypto momentum) better than Sharpe, which penalizes their large favorable moves as if they were risk.
• A smaller sample of downside-only data points makes Sortino noisier than Sharpe when the return history is short.

What Is the Sortino Ratio?

The Sortino ratio isolates the downside from a return stream and prices it separately from the upside.

Developed by Frank Sortino as a refinement of the Sharpe ratio, the Sortino ratio answers a narrower question than Sharpe does: not "how much did returns bounce around in general," but "how much did returns fall short of an acceptable outcome." The formula is:

Sortino Ratio = (Rp − Rf) ÷ σd

Here Rp is the portfolio's average return over the measurement period, Rf is the risk-free rate (or, in many practical implementations, the same minimum acceptable return used to build the denominator), and σd is downside deviation. In practice, most calculators set the numerator's benchmark and the denominator's MAR to the same value — commonly 0%, the current risk-free rate, or a stated target return — so the two don't drift apart, but nothing in the formula forces that. A version that subtracts the risk-free rate in the numerator while measuring downside against a 5% target MAR in the denominator is legitimate and common, but it is a different calculation from one that uses 0% for both, and the two are not directly comparable.

The defining feature is the denominator. A standard deviation, the kind used in the Sharpe ratio, treats a return that is far above average exactly the same as a return that is far below average — both increase the spread, and both increase the "risk" figure. Downside deviation only responds to the second case. A strategy that occasionally posts an enormous gain looks more volatile by the Sharpe ratio's standard-deviation measure, but not by Sortino's, because that gain was never a loss relative to the MAR in the first place.

How Downside Deviation Is Calculated

Downside deviation is a standard deviation computed with every non-loss period treated as a zero.

The calculation proceeds period by period against the chosen MAR:

1. Choose the minimum acceptable return (MAR) for the period being measured — often 0%, the prevailing risk-free rate, or a stated target return such as an 8% annual hurdle.
2. For every period's return, compare it to the MAR. A return at or above the MAR contributes nothing to downside deviation — it is treated as a zero, not skipped or left blank.
3. For every period below the MAR, compute the squared deviation: (return − MAR)². Squaring removes the negative sign and weights larger shortfalls more heavily than small ones.
4. Sum the squared deviations across the entire sample of n periods — including the zeros contributed by periods that were at or above the MAR.
5. Divide that sum by n, the total number of periods in the sample. This is the step most explanations gloss over: the denominator is the full period count, not just the count of periods that fell below the MAR.
6. Take the square root of the result. That figure, σd, is expressed in the same units as the period returns — percent per month, for example — and represents only the downside half of the return distribution's spread.

σd = √( (1/n) × Σ [min(0, Ri − MAR)]² )

Upside volatility is excluded entirely, by construction — a period that beat the MAR by a wide margin adds a zero to the sum, identical to a period that beat it by a narrow margin. Only the shortfall periods, and how far below the MAR each one fell, shape the final number.

A methodology note worth flagging: some calculators and spreadsheets divide the sum of squared shortfalls by the count of shortfall periods only, rather than by the total sample size n. That variant produces a different, typically larger, downside deviation than the standard formula above, because the same sum is spread across fewer periods. Both versions appear in circulation under the same name, which is exactly the kind of inconsistency that makes two "Sortino ratios" quietly incomparable even when the underlying returns are identical — see Common Mistakes below.

Worked Example: Sortino From Eight Monthly Returns

Take eight monthly returns for a hypothetical strategy, with a MAR of 0% — any month with a loss counts as downside.

Four of the eight months — months 2, 4, 6, and 8 — fell below the 0% MAR and count toward downside deviation. The other four, including the two largest moves of the sample (+5.0% and +4.0%), contribute nothing to the denominator no matter how large they were.

Step 1 — average return. Summing all eight returns gives 4 − 2 + 3 − 1 + 5 − 3 + 2 − 1.5 = 6.5, so the average monthly return Rp is 6.5% ÷ 8 = 0.8125%.

Step 2 — downside deviation. The four squared shortfalls sum to 4.00 + 1.00 + 9.00 + 2.25 = 16.25. Dividing by the total sample size (n = 8, not 4) gives 2.03125. The square root of 2.03125 is approximately 1.425%, so σd ≈ 1.425% per month.

Step 3 — monthly Sortino ratio. Using MAR = 0% as Rf as well for consistency, Sortino = 0.8125% ÷ 1.425% ≈ 0.570 on a monthly basis.

Step 4 — annualize. Multiplying the monthly ratio by √12 (≈ 3.464) — the standard annualization factor for monthly data — gives an annualized Sortino ratio of approximately 1.98.

Now compute the Sharpe ratio on the identical eight months, using the full standard deviation rather than downside deviation. The mean deviation of each return from the 0.8125% average, squared and summed across all eight months, totals 64.97; dividing by n = 8 gives a variance of 8.121, and the square root is a total standard deviation of approximately 2.850% per month — almost exactly double σd. The monthly Sharpe ratio is 0.8125% ÷ 2.850% ≈ 0.285, which annualizes (× √12) to approximately 0.99.

Side by side: Sortino ≈ 1.98 versus Sharpe ≈ 0.99 — Sortino comes out almost exactly double, for the same eight returns. The reason is visible directly in the table: the two largest single-month moves in the sample, +5.0% and +4.0%, drag total standard deviation upward in the Sharpe calculation but contribute nothing to downside deviation in the Sortino calculation. Sharpe treats a great month and a terrible month as symmetric contributors to risk; Sortino only counts the terrible ones. Neither ratio is "wrong" — they are answering different questions about the same eight numbers.

When Sortino Is More Appropriate Than Sharpe

Sortino earns its keep specifically when a strategy's return distribution is asymmetric — meaningfully skewed rather than roughly symmetric around the mean.

Options-selling strategies

A premium-selling program — covered calls, cash-secured puts, credit spreads — typically produces a long run of small, fairly uniform gains punctuated by occasional sharp losses when the underlying makes a large move against the position. The small gains barely move a standard-deviation calculation, but a Sharpe ratio still folds whatever mild variance exists among them into the same "risk" bucket as the rare large loss. Sortino separates the two: the steady collection of premium contributes almost nothing to σd, while the tail loss — the outcome that actually threatens the account — is what shapes the ratio.

Trend-following with a fat right tail

Trend-following systems are built to cut small losing trades quickly and let winning trades run. The return distribution that results is lopsided on purpose: many small losses, and a handful of outsized winners that can dwarf everything else in the sample. Those big winners inflate total standard deviation and drag the Sharpe ratio down, penalizing the strategy for the very upside that makes it work. Downside deviation never sees those winning months as anything other than zero contributions, so Sortino credits the strategy for its actual loss-containment discipline without punishing it for large gains.

Crypto momentum strategies

Momentum strategies trading crypto assets are exposed to the same dynamic in a more extreme form — explosive short-term rallies of 50% or more are not unusual, sitting alongside grinding chop or sharp corrections. A Sharpe ratio treats a strategy that captured a 100% rally as symmetrically "risky" to one that lost 100%, which is not a useful way to describe what actually happened to the account. Sortino's asymmetry matches the asset class's own asymmetry.

The general principle behind all three cases: when a return distribution is close to symmetric, Sortino and Sharpe tend to rank strategies similarly, and the choice between them matters less. The more skewed the distribution — in either direction — the more the two ratios can disagree, and the more it matters which question is actually being asked.

Limitations of the Sortino Ratio

Two limitations matter enough to state every time the ratio is used: sample size and the choice of MAR.

A smaller, noisier sample

Downside deviation is built from only the periods that fell below the MAR — in the worked example above, four out of eight months, but real strategies can easily see that fraction run lower. A Sharpe ratio calculated on 36 months of history draws its standard deviation from all 36 data points. A Sortino ratio calculated on the same 36 months might build its downside deviation from ten or twelve losing months. Ten or twelve data points produce a much noisier estimate of the true underlying volatility than thirty-six do — one additional bad month, or the removal of one that happens to roll off a trailing window, can swing σd, and therefore the ratio, by a wider margin than the equivalent change would move a Sharpe ratio built on the full sample. This makes Sortino comparisons between strategies with short track records considerably less reliable than the same comparison run on a longer history.

The MAR choice changes the number

0%, the risk-free rate, and a stated target return are three different thresholds, and each produces a different downside deviation from the identical return series — a period that is a "loss" relative to a 5% hurdle may be a "gain" relative to 0%. Because the MAR also frequently doubles as the benchmark subtracted in the numerator, changing it can shift both the numerator and the denominator at once. Two Sortino ratios computed on the same underlying returns with different MAR assumptions are not the same statistic wearing different clothes — they are different calculations that happen to share a name, and quoting one without stating the MAR used strips the number of the context needed to interpret it.

Misconceptions Versus Reality

Common Mistakes When Using the Sortino Ratio

• Comparing ratios built on different MAR assumptions. A Sortino ratio quoted with a 0% MAR and one quoted with a risk-free-rate MAR are not measuring the same downside, even if both are labeled "Sortino ratio" and computed from similar return histories. Before comparing two published numbers, confirm the MAR used for each.
• Using too short a sample. Because downside deviation draws only from below-MAR periods, a short history — a few months, or even a year of monthly data — can leave the entire calculation resting on three or four data points. A ratio built on that few observations is closer to an anecdote than a statistic, and should be treated with proportionally less confidence than the same-length Sharpe ratio.
• Mixing the total-n and downside-count-n conventions without noticing. As covered above, some tools divide by the total sample size and others divide only by the count of shortfall periods. Pulling one figure from a spreadsheet built one way and another from a data provider built the other way produces an apples-to-oranges comparison that looks apples-to-apples.
• Treating a strong Sortino ratio as proof a strategy is safe. The ratio only reflects the downside that occurred within the measured sample. A strategy that has not yet lived through its worst-case scenario will show a favorable Sortino ratio right up until it does.
• Annualizing inconsistently. Multiplying a monthly ratio by 12 instead of √12, or annualizing the numerator and denominator with different conventions, produces a number that looks plausible but does not correspond to any coherent methodology.

Risks, Limitations, and Exceptions

• Downside deviation reflects only the shortfall periods actually observed in the sample; it cannot anticipate a tail event that has not yet occurred.
• The MAR is a subjective choice, and no single value is universally correct for every investor or strategy.
• A short return history produces a downside deviation estimated from very few data points, making the ratio unstable.
• Different calculators use different denominator conventions (total periods versus downside-period count only), producing numbers that are not directly comparable despite sharing a name.
• Like standard deviation, downside deviation is sensitive to a single large outlier, which can dominate the calculation in a short sample.
• Annualization conventions vary; a ratio calculated from monthly data and one calculated from daily data are not automatically comparable without matching methodology.
• A favorable Sortino ratio does not guarantee a favorable future outcome and does not bound the size of a future loss.
• The ratio measures dispersion of shortfalls, not their magnitude beyond that dispersion — it is still a variance-based measure, not a direct measure of worst-case loss.

Practical Implementation Checklist

1. Choose a MAR (0%, the risk-free rate, or a stated target return) before calculating anything, and document the choice.
2. Gather a return series long enough that the below-MAR subset will contain a meaningful number of data points, not just a handful.
3. Compute each period's shortfall relative to the MAR, squaring only the below-MAR periods and treating the rest as zero.
4. Divide the sum of squared shortfalls by the total number of periods in the sample (n), not just the count of below-MAR periods.
5. Take the square root to get downside deviation, then divide excess return by that figure to get the period Sortino ratio.
6. Annualize using a consistent, stated convention (such as multiplying by √12 for monthly data) and disclose it alongside the number.
7. Report the MAR, the sample length, and the annualization method every time the ratio is quoted or compared to another source.
8. Pair the Sortino ratio with a maximum-drawdown or tail-risk figure rather than relying on it alone, since it cannot anticipate an unobserved tail event.

Sources and Methodology

This guide follows the downside deviation methodology as originally formalized in the academic literature and as commonly taught in professional investment curricula. Key sources include:

• Sortino, F. A. & van der Meer, R., "Downside Risk," Journal of Portfolio Management (1991): introduced the downside-risk framework distinguishing harmful volatility (below a target return) from beneficial volatility (above it), the conceptual basis for the ratio.
• Sortino, F. A. & Price, L. N., "Performance Measurement in a Downside Risk Framework," Journal of Investing (1994): formalized the downside deviation calculation — averaging squared below-target shortfalls over the total sample size — that this guide's worked example follows.
• CFA Institute curriculum, "Portfolio Risk and Return" readings: covers the Sortino ratio alongside the Sharpe ratio as risk-adjusted performance measures within standard investment-management coursework, including guidance on comparing ratios that use different benchmark assumptions.

The worked numeric example on this page is an illustrative calculation built for teaching purposes, not a performance figure for any actual Swoopr tool, fund, or strategy.

Frequently Asked Questions

Conclusion

The Sortino ratio divides excess return by downside deviation only, ignoring upside swings that the Sharpe ratio treats as equivalent risk.

That single change in the denominator makes Sortino the more informative choice for strategies with asymmetric return distributions, and a noisier, more assumption-dependent one for strategies with short histories or ambiguous MAR conventions. Neither ratio replaces the other; each answers a different question about the same underlying returns. Move to the parent performance-metrics hub for the full set of ratios, or to a specific companion page when a related calculation is needed.

Related Reading

• Portfolio performance metrics hub — the parent guide this page belongs to.
• Sharpe ratio — the standard-deviation-based measure Sortino refines by excluding upside volatility.
• Calmar ratio — a drawdown-based alternative that measures return against maximum peak-to-trough loss instead of return volatility.
• Portfolio risk — the broader framework of heat, correlation, concentration, and drawdown that these performance ratios sit alongside.