A Sortino ratio calculator divides excess return by downside deviation rather than by total volatility. The measure is associated with Frank Sortino's work on downside risk in the 1980s and 1990s, and it exists because of a specific objection to the Sharpe ratio: standard deviation treats an unexpectedly good period as risk in exactly the same way it treats an unexpectedly bad one.
Arb Digital publishes this next to the Sharpe ratio calculator, which uses total volatility in the denominator. The boundary is that one line of arithmetic: same numerator, different measure of dispersion. Both describe a period that has already happened, and neither is a recommendation to buy, sell or hold anything.
What This Sortino Ratio Calculator Does
It computes downside deviation properly, which is where most implementations differ from each other.
Downside deviation squares only the shortfalls below a minimum acceptable return, treats every period at or above that threshold as contributing zero, and then divides by the count of all observations rather than only the bad ones. That last detail is the one that varies in the wild, and it changes the answer substantially. This page uses the standard convention by default and offers the alternative under a clearly labelled option so the size of the divergence is visible rather than hidden.
Alongside the ratio it reports the Sharpe ratio on the same data, using the same population convention, so the two are directly comparable. It also reports the annualised Sortino, scaled by the square root of the number of periods per year, and the raw downside deviation so you can see what the denominator actually was.
The threshold is yours. Zero is common, the risk-free rate is common, and a target return is common. All three produce different ratios on identical returns, which is why a Sortino ratio quoted without its threshold is not comparable with anything.
How to Use It
- Use one consistent period. Monthly returns throughout, or quarterly throughout. Mixing frequencies makes both the mean and the deviation meaningless.
- Enter returns as percentages. A 2% month goes in as 2, not 0.02. The ratio is unit-free, so consistency matters more than the scale you choose.
- Set the threshold deliberately. Zero means any loss counts as downside. A target return means any period that misses the target counts, including small positive ones.
- Match the annualisation to the data. Twelve for monthly, four for quarterly, around 252 for daily trading returns.
- Give it enough observations. With a short series only a handful of periods fall below the threshold, and the denominator rests on very little data.
The Formula / How It's Calculated
Downside deviation measures dispersion below the minimum acceptable return, treating everything above it as zero:
DD = √[ ( Σ min(0, Ri − MAR)² ) ÷ N ]
where N is the total number of observations, not the number that fell below the threshold. The ratio is then mean excess return over that deviation:
Sortino = (mean R − MAR) ÷ DD
Annualising scales the per-period figure by the square root of the number of periods in a year, which assumes returns are independent across periods:
Annualised Sortino = Sortino × √(periods per year)
The Sharpe ratio for comparison uses the same numerator structure with total standard deviation in the denominator:
Sharpe = (mean R − risk-free rate) ÷ σ
Worked example, matching the values the page loads with. Twelve monthly returns are supplied, with a threshold of zero. Their mean is 0.683333%. Four of the twelve are negative: −1.5, −2.4, −0.6 and −3.1. Squaring those and summing gives 2.25 + 5.76 + 0.36 + 9.61 = 17.98. Dividing by twelve — all observations, not four — and taking the square root gives a downside deviation of 1.224064.
The ratio is therefore 0.683333 ÷ 1.224064 = 0.5582, and annualised by √12 it is 1.9338. The Sharpe ratio on the same series, with total standard deviation of 2.123611, is 0.3218. The Sortino figure is higher because most of the dispersion in this series is upside, and Sharpe charges for it while Sortino does not.
The Divisor Is Where Implementations Disagree
This is the correctness point, and it is worth being precise about because two calculators can produce different Sortino ratios from identical data without either containing a bug.
The standard definition divides the sum of squared shortfalls by the total number of observations. Every period at or above the threshold contributes a zero to the numerator of that average, and still counts in the denominator. That is deliberate: it means downside deviation falls when a strategy produces more periods above the threshold, which is exactly the behaviour the measure is meant to have.
The variant divides by the count of shortfalls only. That is a conditional deviation — the typical size of a bad period, given that a period was bad — and it is a legitimate statistic answering a different question. Because it divides by a smaller number, it always produces a larger denominator and therefore a smaller ratio.
In the worked example above the difference is not marginal. Dividing 17.98 by four instead of twelve gives a deviation of 2.120142 rather than 1.224064, and the ratio falls from 0.5582 to 0.3223 — a change of more than 40% from a convention. Anyone comparing a Sortino ratio across two sources without checking the divisor is comparing two different statistics.
Why the Threshold Changes Everything
The minimum acceptable return is not a technical detail. It defines what the word downside means in the output.
At a threshold of zero, only losses are penalised, and a period returning 0.1% is treated as entirely satisfactory. At a threshold equal to the risk-free rate, any period that underperforms cash counts as a shortfall. At a threshold set to a target — a required return, a benchmark, a liability the portfolio has to meet — periods that were positive but insufficient are penalised too, which is often the honest framing for a portfolio with an actual obligation to fund.
Because raising the threshold both reduces the numerator and increases the denominator, the ratio falls twice as fast as intuition suggests. A series with a comfortable-looking ratio at zero can look mediocre against a modest target, and neither figure is wrong. They answer different questions.
This asymmetry is also the reason the measure is not simply a friendlier Sharpe ratio. It is a different question, and the empirical case that downside risk is priced separately from total volatility has been studied directly — Ang, Chen and Xing's NBER working paper Downside Risk examines whether stocks that fall harder with the market when the market declines earn different average returns, which is the underlying premise the measure rests on.
What the Ratio Does Not Capture
Sortino is a single number computed from a completed return series, and it inherits every limitation that description implies.
It says nothing about the path. Two series with the same mean and the same downside deviation can have very different sequences, and one may have delivered its losses consecutively in a way that would have forced a liquidation. The max drawdown calculator measures that peak-to-trough path directly, and the value at risk calculator takes the threshold-and-tail view. All three are different windows on the same idea.
It says nothing about the shape of the tail beyond the second moment. Squaring shortfalls penalises large ones more than small ones, but a distribution with rare catastrophic losses can show a modest downside deviation for years before producing one.
It says nothing about liquidity, counterparty exposure, leverage or concentration, none of which appear in a return series at all. FINRA's overview of investment risk sets out the categories a dispersion measure cannot see.
And it says nothing about the future. Every figure on this page describes returns that have already occurred. Past behaviour does not predict future returns, and a high ratio over one window is not evidence about the next one.
Arb Digital reports performance the way this page reports a ratio — the window stated, the threshold named, and the bad periods shown rather than averaged away.
See SEO Services Talk to Arb DigitalCommon Mistakes to Avoid
- Dividing by the count of losses — the standard definition divides by all observations, and using the smaller count can change the ratio by 40% or more.
- Quoting a ratio without its threshold — the same returns produce different figures at zero, at the risk-free rate and at a target, so the number alone is not comparable.
- Mixing frequencies — a series that is part monthly and part quarterly produces a mean and a deviation that describe nothing.
- Comparing an annualised figure with a per-period one — the square-root scaling makes them differ by a factor of more than three for monthly data.
- Reading it as a substitute for drawdown — the ratio is order-independent, so it cannot see a run of consecutive losses that a path-based measure would show immediately.
Related Free Tools From Arb Digital
The Sharpe ratio calculator is the same numerator over total volatility, and the standard deviation calculator produces that denominator on its own. For path and tail views of the same downside, use the max drawdown calculator and the value at risk calculator. The variance calculator handles the squared deviations directly, and the stock return calculator and annualized return calculator build the series in the first place. Everything else is in the free online tools hub.
Frequently Asked Questions
It is mean excess return divided by downside deviation, where downside deviation measures dispersion only below a minimum acceptable return. It was developed as an answer to the objection that standard deviation, used in the Sharpe ratio, treats unusually good periods as risk in the same way it treats unusually bad ones.
Only in the denominator. Sharpe uses total standard deviation, which counts deviations in both directions. Sortino uses downside deviation, which counts only shortfalls below the threshold. For a series whose dispersion is mostly upside, the Sortino figure is the higher of the two.
By all of them, in the standard definition. Periods at or above the threshold contribute zero to the sum of squared shortfalls but still count in the divisor. Dividing by the count of losses alone gives a conditional deviation, which is a different and always larger figure that produces a lower ratio.
Whatever threshold matches the question. Zero penalises only losses, the risk-free rate penalises any period that underperforms cash, and a target return penalises periods that were positive but insufficient. All three are defensible, and the choice has to be stated with the result.
Yes, whenever the mean return falls below the threshold. The magnitude of a negative ratio is difficult to interpret, because a smaller downside deviation makes an already negative number more negative. Comparing negative Sortino ratios across portfolios is not reliable.
Multiply the per-period ratio by the square root of the number of periods in a year — √12 for monthly data, √252 for daily. That scaling assumes returns are independent from one period to the next, which real return series often violate through volatility clustering.
Downside deviation is zero and the ratio is undefined, because the calculation divides by it. That is not evidence of a risk-free series; it means the sample contained no shortfalls, which is far more likely in a short window than a long one.
It means that over the specific window measured, and against the specific threshold chosen, return per unit of downside deviation was higher. It describes past returns only. Past behaviour does not predict future returns, and the ratio ignores liquidity, leverage, concentration and the order in which the returns arrived.
This tool performs a statistical calculation on data you supply. It is not investment advice, not a recommendation to buy, sell or hold any security, and its output is not a forecast. Measures computed from past returns describe past returns only. Decisions about your money should involve a licensed financial adviser regulated in your jurisdiction.
