Sharpe, Sortino and Risk-Adjusted Return
Return per unit of risk, where risk means the variability of returns. The measures are useful, widely quoted, and rest on assumptions that returns do not satisfy.
MadStockAlerts Research · Updated August 29, 2026
What to take away
- Sharpe is excess return divided by the standard deviation of returns.
- Sortino uses only downside deviation, on the view that upside variability is not risk.
- Both assume a distribution that real returns do not follow.
- A strategy that sells options can post an excellent ratio and carry a large tail risk.
- Ratios are comparable only over the same period and the same frequency.
MAD Academy Training Video · 0:45
Return Per Unit of Discomfort
Sharpe divides excess return by total volatility; Sortino divides it by downside volatility only, which changes who looks good.
This lesson is part of a Stock Alerts + Tools plan.
The two measures
Sharpe = (portfolio return - risk-free rate) / standard deviation of returns
- the numerator is the return earned above holding cash
- the denominator treats upside and downside variability identically
Sortino = (portfolio return - target return) / downside deviation
- downside deviation counts only periods below the target
- the reasoning is that variability to the upside is not what anyone means by risk
The difference matters most for portfolios with asymmetric returns. A strategy with occasional very large gains is penalised by Sharpe for those gains and not by Sortino, which is why the two can rank the same set of portfolios differently.
What the assumptions require
- That standard deviation captures risk, which requires returns to be approximately normally distributed.
- That returns are independent from period to period, which volatility clustering violates.
- That the risk-free rate is a meaningful reference, which it is, and which changes over time.
- That the period measured is representative, which it may not be.
Real return distributions have fat tails: extreme outcomes occur far more often than a normal distribution implies. A measure built on standard deviation systematically understates the risk of exactly the events that matter most.
The strategy that games it
There is a well-documented way to produce an excellent Sharpe ratio: collect small, steady premiums with a small probability of a very large loss. Selling insurance-like exposure does this, and so does any strategy that is short volatility.
Until the loss occurs, the return series is smooth, the standard deviation is low, and the ratio is outstanding. The risk is entirely in a tail that the measure does not see, and the measure is not wrong about the data it was given.
This is why a very high ratio over a short period warrants more scrutiny rather than less. The question is what would have to happen for the strategy to lose a great deal at once, and no risk-adjusted return measure asks it.
Scroll the chart sideways to see all of it.
Using them fairly
- Compare only over identical periods. A ratio from one decade and one from another are not comparable.
- Compare only at the same frequency. Monthly and daily calculations produce different numbers for the same portfolio.
- Look at maximum drawdown alongside, since it describes the worst realised experience rather than an average variability.
- Treat any period without a substantial decline in it as an incomplete test.
The third item is the most useful complement. Drawdown is a realised fact rather than a distributional assumption, and a strategy with an excellent ratio and a severe drawdown has told you something the ratio alone did not.
Reading a ratio at face value
A handful of checks separate a ratio that describes something from one that describes a period.
| Check | Why |
|---|---|
| Does the period include a substantial decline | A ratio from an uninterrupted rise has not been tested |
| What is the maximum drawdown alongside it | A realised worst case, rather than an average variability |
| How long is the record | Ratios from short periods are dominated by noise |
| What is the return distribution's shape | Negative skew is invisible in the ratio and decisive in the outcome |
| What could cause a very large loss | The question no risk-adjusted measure asks |
The last row is the one that matters most and is the least quantitative. Every strategy with an exceptional ratio and a hidden tail passes the first four checks right up until the tail arrives.