A high Sortino ratio does not mean a strategy is safe for a prop account. Sortino measures return above a chosen target per unit of downside deviation; it does not measure the probability of breaching a daily or overall loss rule.
That makes it useful, not sufficient. The ratio removes one weakness of Sharpe while leaving path risk, target choice, and sampling risk firmly on the desk.
What the Sortino ratio measures
The Sortino ratio is average return above a minimum acceptable return divided by downside deviation. Unlike Sharpe, it does not penalise returns merely for being far above the mean.
Sortino = (mean return − target return) / downside deviation
downside deviation =
square root of [sum of min(0, return − target)² / all observations]The target return is often called the minimum acceptable return. It defines which outcomes count as shortfalls. A target of flat performance asks a different question from a positive hurdle, so two otherwise identical reports can produce different Sortino ratios.
The CFA Institute explanation of the Sortino ratio makes two details explicit: the target belongs in both the numerator and the risk definition, and squared shortfalls should be divided by the total number of return observations, not only the observations below target.
That denominator detail matters. Dividing only by losing periods makes downside deviation look smaller when losses are rare, which makes the ratio look better.
Why Sortino can look better than Sharpe
Sortino can rank a positively skewed strategy more fairly because large gains do not inflate the risk denominator. Sharpe treats upside and downside dispersion alike; Sortino asks only how returns behave below the chosen target.
That is a defensible distinction. A large positive surprise is not the same account problem as a large negative surprise.
But a higher Sortino than Sharpe is not evidence that downside risk is controlled. It may simply show that much of the strategy's variability was on the upside. The ratio still compresses a complete return path into one numerator and one denominator.
This is why the basis must be labelled. Daily returns and per-trade returns are different series, even when they come from the same trades. The per-trade versus per-day metrics guide explains how changing the observation unit changes the statistic.
The target quietly controls the result
The minimum acceptable return is not a harmless input. It defines failure for the calculation.
If the target is set to flat performance, only negative observations create shortfall. If it is set higher, small positive observations can also enter downside deviation. Neither choice is universally correct; each represents a different objective.
Comparisons therefore require the same target and the same return frequency. Comparing two Sortino ratios calculated with different targets is comparing two different questions.
The CFA Institute paper also warns that several downside-deviation methods exist and can materially alter the result. A clean report must expose the convention instead of printing “Sortino” as if every platform meant the same thing.
What Sortino misses in a prop backtest
Sortino does not encode prop-firm stopping rules. A strategy can post an attractive ratio and still place several losses inside one daily window, carry floating losses through a reset, or touch an overall equity floor.
FTMO's current Trading Objectives illustrate why the distinction matters. Its account variants define Maximum Daily Loss and Maximum Loss using account equity, including open positions, swaps, and commissions. Exact rules and variants can change, so the firm's page belongs in the test specification.
Sortino misses several account-level questions:
| Missing question | Why it matters |
|---|---|
| When did the losses occur? | Loss clustering can breach a daily rule |
| What was open equity doing? | Closed-period returns can hide intraperiod pressure |
| How close was the loss boundary? | The ratio has no concept of a stopping barrier |
| Were losses serially dependent? | Smooth averages can understate clustered downside |
| How many downside observations exist? | A small denominator sample is unstable |
| Was the metric selected after many trials? | Search bias can promote the luckiest ratio |
Autocorrelation is especially awkward. If losses arrive in clusters, downside observations are not independent pieces of evidence. Autocorrelation in trading backtests shows why nominal sample size can overstate statistical information.
How to audit a Sortino ratio
A useful Sortino audit begins by reproducing the calculation, then moves beyond it. The aim is to understand exactly what the ratio summarises and what it leaves out.
Check the following:
- State the return series: per trade, daily, or another fixed interval.
- State the target return in the same units and frequency.
- Show the downside-deviation formula and its observation count.
- Recalculate the ratio without annualisation before examining any scaled version.
- Inspect the ordered equity path, floating drawdown, and rule breaches separately.
- Repeat on untouched out-of-sample data and stress clustered losses with a block bootstrap.
Do not use Sortino as the strategy-selection objective and then cite the same maximum Sortino as independent proof. That is selection bias. A metric chosen by the search needs untouched evidence after the search.
realbacktesting is a trading-software studio for cTrader built around verifiable testing. Its methodology page documents the execution and validation assumptions, while the funding page puts results inside account constraints. A reproducible report lets the reader check the ratio and the path. It does not turn either into a live guarantee.
What this ratio cannot prove
Sortino cannot prove that a strategy's downside distribution will persist. It describes historical returns under a chosen convention.
It also cannot replace drawdown, tail, dependency, or rule-path analysis. Two strategies can share a Sortino ratio while having very different worst losses and very different sequences. The compression is the feature and the limitation.
Frequently asked
Is a higher Sortino ratio always better?
A higher Sortino indicates more return above the chosen target per unit of measured downside deviation, provided the calculations use the same target, frequency, and convention. It does not automatically mean lower prop-rule breach risk.
What is a good Sortino ratio for a prop strategy?
There is no universal threshold because the result depends on the target, observation frequency, sample, and calculation method. A defensible report states those inputs and compares like with like.
Should Sortino replace Sharpe?
No. Sortino answers a narrower downside question, while Sharpe describes total return variability. Both need path, drawdown, and out-of-sample evidence around them.
The stubborn takeaway
Sortino can ignore good volatility. A prop rule will not ignore bad timing.