Quantis Trade
← Back to all metrics

Metrics / Sortino

Sortino Ratio

3 min read · Metric

Sharpe with the upside volatility removed. Useful for asymmetric strategies where high upside vol shouldn't count as risk.

What it is

Sortino is Sharpe's cousin that only counts downside deviation as risk. The intuition: nobody cares about “risk” on the days the strategy ripped higher. Penalising those moves in the denominator (as Sharpe does) is just punishing good outcomes.

Sortino is therefore always ≥ Sharpe by construction, and the gap between them measures how asymmetric the return distribution is. A symmetric distribution has Sortino almost equal to Sharpe; a positively-skewed strategy (lots of small losses, occasional big wins) has Sortino well above Sharpe.

Formula

# Downside deviation: stddev of returns below a target (usually 0 or risk-free)
downside_returns = [r for r in returns if r < target]
downside_dev     = sqrt(mean(r ^ 2 for r in downside_returns))

# Annualised
sortino = (mean(returns) - r_f) / downside_dev * sqrt(periods_per_year)

Note the squaring is over only the negative returns — this differs from a naive “std of negatives only” calculation, which would mishandle the centring.

Typical ranges

  • SPY long-run: 0.6–0.8 (Sharpe 0.4–0.6 × ~1.4).
  • Symmetric strategies: Sortino ≈ 1.3× Sharpe.
  • Positive-skew strategies (trend-following, tail-hedge): Sortino can be 2× Sharpe or more.
  • Negative-skew strategies (option-selling, mean-reversion with no stops): Sortino barely above Sharpe — the upside removal doesn't help because there isn't much upside vol to remove.

Common mistakes

  • Quoting Sortino without Sharpe. The ratio in isolation is hard to anchor. The Sortino-to-Sharpe ratio is the real information — that's where the asymmetry hides.
  • Comparing Sortinos across strategies with different skew. A 2.0 Sortino on a positive-skew strategy is not equivalent to a 2.0 Sortino on a symmetric one — the first benefits from the asymmetry, the second is a stronger raw result.
  • Treating Sortino as drawdown-aware. It isn't — it counts the magnitude of negative returns but not their persistence. Calmar is the drawdown-aware metric.

What the platform flags

Quantis shows Sortino alongside Sharpe so the gap is immediate. If you see Sharpe 0.9 with Sortino 1.1 (gap small), the strategy is roughly symmetric and either metric tells you the same thing. If Sharpe 0.9 with Sortino 1.8 (gap large), the strategy has meaningful positive skew worth understanding before deployment.

Further reading