What Is the Sortino Ratio?
Michael Whitman
·
Trading since 2017, full-time since 2020
·
YouTube creator, 137K subscribers
·
Published 2026-09-14
·
Updated
Sortino ratio divides excess return by downside deviation rather than by total volatility, penalizing only the returns that fell below a target. It corrects the Sharpe ratio's treatment of upside movement as risk, at the cost of estimating a denominator from roughly half the observations.
The Sharpe ratio treats a sharp gain as risk. The Sortino ratio does not, which is obviously more sensible and comes with a statistical cost most people never mention.
How it works
The numerator is the same as the Sharpe ratio — return above a stated target, often a cash rate.
The denominator changes. Instead of the standard deviation of all returns, it uses the deviation of only those below the target.
Returns above the target contribute zero. A month that gained 8% adds nothing to measured risk, which matches how anybody actually experiences it.
Where the Sharpe ratio genuinely goes wrong
Two strategies with identical standard deviation. One produces its variation through occasional large gains; the other through occasional large losses.
The Sharpe ratio scores them the same. It cannot distinguish them, because standard deviation is symmetric and does not know which direction a deviation went.
The Sortino ratio separates them immediately, which is the whole reason it exists and is a real improvement for any strategy with asymmetric returns.
A worked example
Strategy A returns 1% most months with occasional 10% gains. Strategy B returns 2% most months with occasional 8% losses.
Both might show identical standard deviation and therefore similar Sharpe ratios, despite being completely different experiences to hold.
The Sortino ratio scores A far higher. Its deviations are almost entirely upward, so its downside deviation is small and its ratio is large.
And B’s ratio collapses, correctly, because all of its variation is the kind that hurts.
The statistical cost
Halving the sample doubles the noise. If a five-year monthly record has 60 observations and 25 are below target, the denominator rests on 25 numbers rather than 60.
Which makes the ratio unstable. Small changes in the sample period produce larger swings in the Sortino ratio than in the Sharpe ratio computed on the same data.
And the target choice matters. Using zero, a cash rate, or a required return produces meaningfully different results, and the choice is rarely disclosed alongside the figure.
So a reported Sortino ratio is not comparable across sources unless the target and the period match, which is more often assumed than checked — and it is the reason the Sharpe ratio, for all its faults, remains the standard comparison.
The original data
Of every S&P 500 trading day since 1950, 93% closed below the previous record, the worst fall was 56.8%, from 2007 to 2009, and the longest below-peak stretch ran 7.5 years, from 1973 to 1980, before new records followed.
That 93% figure is the point. Being below a prior peak is the normal state of an asset that rises over time, and a measure counting only downside deviation is measuring something that occurs almost constantly rather than rarely.
And day-to-day direction on the index is close to balanced: 53.1% of sessions closed up. Down days are nearly as common as up days, so the downside deviation is built from a large sample rather than a handful of observations, which is the condition under which the calculation behaves reasonably; a strategy whose losses are rare and clustered makes it far less stable.
Where it is most and least useful
Most useful for asymmetric strategies. Trend following, long-volatility positions and anything with occasional large gains are systematically misjudged by the Sharpe ratio and fairly judged by this one.
Least useful for option-selling strategies. They produce steady small gains and rare large losses, so in any period without a loss the downside deviation is tiny and the ratio is enormous.
Which is the dangerous case. A strategy that has not yet had its bad event shows a spectacular Sortino ratio precisely because the risk has not appeared in the sample.
And that is a general warning about all of these measures. Every risk-adjusted ratio is a statement about the observed period, and the strategies that score best are frequently the ones whose risk has not yet been observed.
How it is actually calculated
Downside deviation is not simply the standard deviation of negative returns. Returns above the target are set to zero rather than excluded, and the sum of squared shortfalls is divided by the total number of observations.
That distinction matters. Dividing by only the count of losing periods produces a much larger denominator and a much smaller ratio, and both conventions appear in published figures.
Which is one more reason two reported Sortino ratios may not be comparable. The target, the period and the divisor all vary, and none of them is usually stated alongside the number.
The practical response is to calculate it yourself where the underlying returns are available, and to treat a headline figure from a marketing document as an approximate claim rather than a measurement.
How this connects
Option selling is where the flaw above bites hardest. Selling puts, a short strangle or an iron condor all earn the volatility risk premium, a real source of return, and the ratio records that premium long before it records the tail risk being paid for.
The ulcer index measures what neither ratio contains. It captures how deep and how long an account sits below its previous peak, which is information about time spent below a high that a single denominator cannot hold.
Risk of ruin asks the question behind the denominator. The ratio describes return per unit of downside; risk of ruin asks whether a run of losses can end the account, and that depends on position size rather than on the ratio.
It is usually computed from a backtest. That means overfitting inflates it just as it inflates any other result, and walk forward testing on unseen data is the check on whether the figure survives.
When the target is the cash rate, the treasury bill rate for the same period is the usual choice, and it belongs next to the number.
And expectancy is the trade-level counterpart. Expectancy says whether the average trade pays; this ratio says how uncomfortable the losing periods were along the way.
When it fails
The characteristic failure is a short record with no losses in it. A strategy runs for two years without a meaningful down month, the downside deviation is close to zero, and the Sortino ratio is enormous — occasionally reported as an infinite or undefined figure. It reads as extraordinary skill and is a statement that the sample contains no adverse observations. Option sellers, credit strategies and carry trades all produce exactly this pattern in their good periods, and the ratio is most flattering immediately before the event it cannot see.
A second failure is comparing figures with different targets. The choice of threshold changes the result substantially.
A third is ignoring the smaller sample, which makes the estimate noisier than the Sharpe ratio’s.
A fourth is using it on too short a record, where a handful of observations drive the denominator.
The question: A Sortino of 4.0. Skill or a quiet period?
And a fifth is treating a high ratio as low risk. It means downside has not appeared in the sample, which is not the same as its absence.
Related
Sharpe ratio covers the measure this corrects. Downside risk covers the quantity in the denominator. And Calmar ratio covers the version built on maximum drawdown instead.
This ratio fixes something real. Nobody has ever complained that an investment went up too fast, and the Sharpe ratio penalizes exactly that. The catch is that halving your sample to fix it makes the number less reliable, which is a trade rather than an improvement.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.