What is the Sortino Ratio?
Not all volatility is bad. The Sortino Ratio only penalises the kind that is
When evaluating trading models, not all volatility deserves the same treatment.
A model that occasionally produces a larger-than-expected win is very different from a model that occasionally produces a larger-than-expected loss. The Sortino Ratio is built around exactly this distinction.
The problem with treating all volatility equally
The Sharpe Ratio divides returns by total standard deviation — meaning it treats every deviation from the average equally, whether that deviation is a loss or an unexpected gain.
This is a reasonable simplification in many contexts. But it has a practical limitation: it can make a model with controlled losses and occasional large winners look worse than a smoother model with lower upside.
The Sortino Ratio addresses this by only measuring downside deviation — the variability of returns that fall below a target threshold. Upward deviations are simply not counted against the model.
How the Sortino Ratio is calculated
The basic form of the Sortino Ratio is:
Sortino = (Mean Return − Target Return) / Downside Deviation
The target return is typically zero or a minimum acceptable return. Downside deviation is the standard deviation calculated only from returns that fall below that target.
A higher Sortino Ratio means the model generates more return per unit of harmful downside risk. A low or negative Sortino Ratio suggests the downside behaviour is disproportionate relative to what the model is producing on the upside.
How it differs from the Sharpe Ratio in practice
Both ratios describe risk-adjusted quality, but they answer slightly different questions.
The Sharpe Ratio asks: how consistent is the return stream overall? It penalises any kind of variability, positive or negative.
The Sortino Ratio asks: how well does the return stream hold up specifically on the downside? It tolerates upside variability while remaining focused on loss behaviour.
For trading models with asymmetric return profiles — where large wins occur alongside controlled losses — the Sortino Ratio often provides a more meaningful picture than the Sharpe Ratio alone. A model that runs winners and cuts losers may show a higher Sortino than Sharpe, which is often exactly the intended characteristic.
How darwintIQ uses the Sortino Ratio
In darwintIQ, the Sortino Ratio is one of the metrics available in the Trading Model's Stability Layer. It is shown alongside other risk-adjusted measures including the Sharpe Ratio, Calmar Ratio, and Profit Factor.
No single metric tells the full story of a model's quality. The Sortino Ratio contributes one specific lens: how well does the model manage its losing behaviour relative to what it generates on the winning side.
When evaluating two models with similar overall returns, a meaningfully higher Sortino Ratio on one of them is a signal worth paying attention to. It suggests the model's downside is more controlled — which, in changing market conditions, is often a more durable quality than raw profitability alone.
As with all metrics in darwintIQ, the Sortino Ratio is calculated on the current rolling evaluation window. It reflects recent behaviour, not a long historical average.
Final thoughts
The Sortino Ratio is a more precise version of risk-adjusted return measurement. By focusing only on downside deviation, it avoids penalising models for producing large wins and keeps the attention where it belongs: on how losses behave. Used alongside Sharpe and Drawdown metrics, it gives a more complete picture of how a trading model actually performs under pressure.