The Sharpe ratio measures average return above a risk-free benchmark relative to total return variability. The Sortino ratio measures average return above a chosen target relative to shortfalls below that target. Their main difference is what counts as risk: Sharpe includes variability in both directions; Sortino focuses on returns below a threshold.
That distinction can change how you rank two funds, portfolios or trading strategies. Neither ratio replaces a broader investment performance assessment. A useful comparison starts with consistent data, clear calculation rules and an explanation of the risks behind the numbers.
Sharpe vs Sortino: What Each Ratio Measures
Sharpe asks: how much excess return did the investment produce for each unit of variability? Sortino asks: how much return above my target did it produce for each unit of downside deviation?
“Downside” does not necessarily mean a negative return. If your monthly target is 1%, a return of 0.5% falls below it, despite making money. With a zero target, only negative returns contribute to downside deviation.
The distinction is mathematical, not a contest between a good measure and a bad one. Total variability and target shortfalls answer different questions. Before choosing either ratio, decide which question matters for the comparison.
How to Calculate the Sharpe Ratio
For historical observations, calculate each period’s excess return by subtracting its risk-free return from the portfolio return. Divide the arithmetic average of those excess returns by their standard deviation. This differential-return formulation is set out in William Sharpe’s paper on the Sharpe ratio.
Sharpe ratio = Average periodic excess return ÷ Standard deviation of periodic excess returns
When the risk-free return is constant throughout the sample, subtracting it does not change standard deviation. The formula then becomes average portfolio return minus the risk-free return, divided by portfolio return standard deviation.
Suppose average monthly return is 1%, the monthly risk-free return is 0.2%, and monthly standard deviation is 2%. The monthly Sharpe ratio is:
(1% − 0.2%) ÷ 2% = 0.40
Standard deviation measures departures from the mean in either direction. An unusually profitable month contributes to variability, just as an unusually poor month does. But a large gain also increases average return, so it does not automatically reduce Sharpe. Both parts of the fraction change.
How to Calculate the Sortino Ratio
The target-based Sortino ratio replaces the risk-free benchmark with a minimum acceptable return, often abbreviated to MAR. It also replaces total standard deviation with downside deviation. Some published versions retain a risk-free numerator, so check the definition before comparing figures; these variations appear in the CAIA primer’s discussion of risk-adjusted performance measures.
This article uses the same target in both parts of the calculation:
Sortino ratio = (Average periodic return − Target return) ÷ Downside deviation
For a constant target, T, and N observations, calculate historical downside deviation as:
Downside deviation = √{Σ[min(0, rt − T)]2 ÷ N}
Subtract the target from every return. Keep negative differences and replace positive differences with zero. Square those values, average them across all observations, then take the square root.
This is not the standard deviation of losing months. Those are different calculations. Dividing by only the number of below-target observations also changes the measure: it removes the effect of how frequently shortfalls occur.
Choose the Target Before Comparing Results
A zero target answers a different question from a cash benchmark or a required return. Set the threshold before examining which choice produces the most flattering result.
Keep the target’s period consistent with the observations. If an annual target of 6% represents an effective compounded return, its equivalent monthly rate is:
Monthly target = (1.06)1/12 − 1 ≈ 0.4868%
Dividing 6% by 12 gives 0.5%, a simple-rate convention rather than the exact compounded equivalent. Either approach needs a clear label. A monthly threshold also measures monthly shortfalls; it does not directly measure failure to reach an annual goal after intervening gains and losses.
Worked Example: The Rankings Can Reverse
Consider two hypothetical strategies over four months. Assume both the risk-free return and the Sortino target are zero. All figures below are monthly and unannualized. Four observations make the arithmetic easy to inspect, not a sufficient performance record.
| Observation or measure | Strategy A | Strategy B |
|---|---|---|
| Month 1 | −1% | −2% |
| Month 2 | 1% | 2% |
| Month 3 | 1% | 4% |
| Month 4 | 7% | 4% |
| Arithmetic average return | 2.00% | 2.00% |
| Sample standard deviation | 3.464% | 2.828% |
| Downside deviation | 0.50% | 1.00% |
| Sharpe ratio | 0.577 | 0.707 |
| Sortino ratio | 4.00 | 2.00 |
Both strategies have a 2% arithmetic average monthly return. Strategy B has less total variability, so it has the higher Sharpe ratio. Strategy A has the smaller below-zero shortfall, so it has the higher Sortino ratio.
For Strategy A, the deviations from its 2% mean are −3, −1, −1 and 5 percentage points. Using sample standard deviation, which divides squared deviations by N − 1:
Standard deviation A = √[(9 + 1 + 1 + 25) ÷ 3] ≈ 3.464 percentage points
Its only below-target return is −1%. The other three months contribute zero to downside deviation:
Downside deviation A = √[(1 + 0 + 0 + 0) ÷ 4] = 0.50 percentage points
Strategy B’s downside deviation is √(4 ÷ 4), or 1 percentage point. Its larger loss creates a larger downside denominator, even though its returns cluster more closely around the mean.
The ranking reversal follows from the definitions. Strategy A’s 7% month raises its total variability but contributes nothing to downside deviation at a zero target. Sharpe favors B’s steadier sample; Sortino favors A’s smaller observed shortfall. Neither result establishes which strategy will perform better next month.
Annualization: Keep the Units Consistent
A monthly ratio cannot be compared directly with an annualized one. Under the usual assumptions of stable return characteristics, uncorrelated excess returns and additive aggregation, Sharpe is commonly annualized using:
Annualized Sharpe ≈ Monthly Sharpe × √12
The earlier monthly Sharpe of 0.40 becomes approximately 1.39. Compounding complicates the relationship, and serial correlation can make square-root scaling misleading. Positive serial correlation can cause the conventional calculation to overstate annual Sharpe, a problem examined in Andrew Lo’s research on Sharpe ratio estimation and time aggregation.
Sortino requires separate care. Multiplying monthly downside deviation by √12 does not generally reproduce downside deviation calculated from annual returns. A losing month may be offset by gains elsewhere in the year. The threshold is applied to a different distribution at each frequency; the CFA Institute discussion of Sortino calculation pitfalls addresses this problem.
Keep arithmetic averages separate from compounded growth too. Substituting CAGR into a numerator while retaining volatility estimated from monthly arithmetic returns creates a different calculation. The distinction between CAGR and arithmetic average returns matters here: one describes compounded wealth growth, the other averages periodic outcomes.
How to Interpret High, Low and Negative Ratios
With positive numerators and matching calculation settings, a higher ratio indicates more measured reward per unit of the chosen risk measure. It does not establish that an investment suits a particular investor.
A negative Sharpe means average return fell below the selected risk-free benchmark. A negative target-based Sortino means average return fell below the target. Negative ratios need caution: holding a negative numerator constant while increasing the denominator moves the ratio closer to zero. A numerically higher result can therefore arise from more variability.
A Sortino of 2 is not automatically better than a Sharpe of 1. They have different denominators and may have different numerators. Compare Sharpe with Sharpe, and Sortino with Sortino, using matching settings.
If no observation falls below the target, historical downside deviation is zero and the Sortino division is undefined. A reporting system might display infinity or leave the field blank. Neither display proves an absence of future losses. The denominator has run out of observations, not risk.
Where Both Ratios Can Mislead
They Do Not Describe the Order of Losses
Rearrange a fixed set of periodic returns and both ratios remain unchanged, assuming the same constant benchmarks. Mean, standard deviation and downside deviation do not depend on observation order.
The experience of holding the investment can still change. Consecutive losses may create a deeper decline than losses separated by recoveries. Review maximum drawdown and peak-to-trough losses alongside either ratio rather than treating downside deviation as a substitute.
The Inputs Deserve as Much Attention as the Formula
Use the same dates, return frequency, currency and treatment of costs when comparing investments. For trading strategies, measure returns on a consistent account-equity basis, including open positions. A series containing only closed winners while losing positions remain open is not a useful input.
Inspect the underlying returns too. Ask how much of the result depends on the best month, how many observations fall below the target, and whether the sample includes the conditions the strategy is designed to survive. Recalculate after removing an unusually strong observation as a sensitivity check, not as an excuse to discard inconvenient data.
For researched strategies, compare development results with walk-forward and out-of-sample performance. A ratio calculated after repeatedly adjusting a strategy to historical results deserves a different level of confidence from an unchanged rule tested on unseen observations.
Which Ratio Should You Use?
Use Sharpe when the comparison concerns excess return relative to total variability. Add Sortino when returns below a stated threshold are central to the decision. Report both when their disagreement helps explain the return pattern, as it does in the worked example.
For portfolio decisions, also examine how the investment interacts with existing holdings. A standalone ratio does not answer a portfolio correlation and covariance question.
The most useful report shows the ratio, its numerator, its denominator and its calculation settings. Sharpe summarizes reward relative to variability; Sortino summarizes reward relative to target shortfalls. Choose the measure that fits the question, then inspect what the summary leaves out.