Sortino ratio

The Sortino ratio answers the same question as the Sharpe ratio with one correction: it stops treating gains as risk.

Sharpe divides your excess return by total volatility, which counts every move away from average, including the ones in your favour. A portfolio that jumped sharply upward is penalised exactly as much as one that fell just as sharply. That does not match how anyone actually experiences risk. Sortino fixes it by measuring only the falls.

How it is calculated

Sortino ratio = (mean daily return − daily risk-free rate) / downside deviation × √252

The numerator is identical to Sharpe. What changes is the denominator.

Downside deviation uses only the days your portfolio fell. Gylder squares each negative daily return, averages those squares across the number of down days, and takes the square root. Days you gained contribute nothing at all, neither to the sum nor to the count.

The result is annualised with √252, the same as everything else in the risk grid.

The risk-free rate is an assumption

Gylder currently uses a fixed 3% annual risk-free rate, not a live market rate. It is a configuration constant chosen to sit roughly in the euro short rate environment, and it is disclosed here because it feeds the number.

This matters when you compare your figure against one from somewhere else. A Sortino ratio computed against a different risk-free assumption is not directly comparable to this one, and the gap widens the lower your returns are.

How to read it

Higher means more excess return per unit of downside movement. Because the denominator ignores upside, a Sortino ratio is normally higher than the Sharpe ratio for the same portfolio and window.

That gap is the interesting part. A portfolio whose Sortino sits far above its Sharpe is one whose volatility is mostly upward, which is the pleasant kind. A portfolio where the two are close has volatility spread evenly in both directions.

Gylder does not tell you what a good Sortino ratio is. Any threshold we picked would be an opinion presented as a fact, and it would depend on the asset class and the window in ways a single number cannot carry.

What it misses

Dividing by the count of down days only means a portfolio with very few losing days can produce a small denominator and therefore a large ratio, on thin evidence. Treat a striking figure over a short window as a small sample rather than a finding.

It also inherits Sharpe's blind spots. It says nothing about how the losses were distributed in time, so it cannot distinguish many small falls from one severe one. For that, see maximum drawdown and the Calmar ratio.