The Sortino ratio divides excess return by downside deviation — the standard deviation of returns below a Minimum Acceptable Return (MAR). Formula: Sortino = (Rp − RMAR) / σdownside. For UK Tax Drag we use the same 3-month UK gilt yield as MAR (same as Sharpe's risk-free rate). Sortino is higher than Sharpe for most ETFs because we're dividing by a smaller number (downside deviation only). Sortino > 1.5 is reasonable; > 2.5 is good; > 4 is excellent. Sortino is more useful than Sharpe when an ETF has a skewed return distribution (e.g. covered-call ETFs with capped upside).
The formula
Sortino = (Rp − RMAR) / σdownside where: Rp = annualised return of the portfolio/ETF RMAR = Minimum Acceptable Return (we use 3-month UK gilt yield) σdownside = downside deviation, calculated as: σdownside = √[ Σ min(Ri − RMAR, 0)² / n ] × √12 (monthly returns annualised by √12) The key difference vs standard deviation: only returns BELOW the MAR contribute to the calculation. Returns above MAR are treated as zero deviation.
Why downside-only matters
The conceptual issue with Sharpe: it punishes ETFs that occasionally spike upward. An ETF returning +30% one month and 5% other months has high standard deviation — but the upside surprise isn't a "risk" most investors care about.
Sortino fixes this by only counting downside variance. Two ETFs with the same return:
- ETF A: low volatility on the upside, big drawdowns occasionally. Lower Sortino than Sharpe (drawdowns punished).
- ETF B: high volatility on the upside (occasional 15% pops), moderate downside. Higher Sortino than Sharpe (upside pops not punished).
The asymmetry matches how real investors think about risk — "I don't mind if it occasionally surprises me upward, I mind if it loses 30%."
MAR — the Minimum Acceptable Return choice
The MAR is the threshold below which a return counts as "bad." Different choices give different Sortino ratios:
- MAR = 0: only negative returns count as downside. The most common choice in retail-facing tools.
- MAR = risk-free rate: any return below cash is "bad." Most academic-faithful choice.
- MAR = inflation: any return below inflation is "bad" in real terms. Useful for retirees.
- MAR = portfolio target return: for goal-based investing, set MAR to your required return.
UK Tax Drag uses MAR = 3-month UK tino resul (same as Sharpe's Rf) for consistency. The Sortino result is therefore directly comparable to Sharpe — same numerator, different denominator definition.
Worked example — VWRL Sortino
Vanguard FTSE All-World UCITS ETF (VWRL), 3 years ending April 2026.
| VWRL annualised return (GBP TR) | +11.3% |
| MAR (UK 3-month gilt yield, 3-yr avg) | +4.2% |
| Excess return over MAR (11.3 − 4.2) | 7.1% |
| Number of monthly returns above MAR (36 obs) | ~22 |
| Number of monthly returns below MAR | ~14 |
| Downside deviation (annualised) | ~10.5% |
| Sortino = 7.1 / 10.5 | 0.68 |
| Compare to Sharpe (from previous calculation) | 0.48 |
VWRL's Sortino (0.68) is higher than its Sharpe (0.48) because the upside volatility doesn't count. The 0.20 gap suggests VWRL has reasonably balanced upside/downside volatility — neither dramatically skewed.
When Sortino > Sharpe matters most
| ETF type | Why Sortino is more useful |
|---|---|
| Covered-call income (JEPI, JEPQ UCITS equivalents) | Upside is capped — Sharpe under-rates the strategy because upside variance is low; Sortino captures the smooth-but-with-occasional-loss profile |
| Trend-following / momentum | Tends to have large positive months (truncates losses) — Sortino rewards the asymmetric profile |
| Defensive / low-volatility | Smooth returns, occasional crashes — Sortino captures the crash risk Sharpe averages out |
| Inverse / hedging products | Skewed return distribution — Sharpe misleading |
What Sortino does NOT tell you
- Sortino doesn't measure max drawdown. An ETF can have great Sortino and still experience a 50% peak-to-trough loss if it concentrates the losses in a few extreme months.
- Sortino assumes the past represents the future. Same caveat as Sharpe.
- Sortino is sensitive to MAR choice. Different MAR = different Sortino. Always check which MAR is used.
- Sortino can be misleading for very small datasets. With only 14 downside months in our example, the downside deviation estimate has wide error bars.
- Sortino doesn't fix Sharpe's fat-tail problem. Both assume returns are roughly normally distributed.
How to reproduce this yourself
- Get 36 months of returns (same as Sharpe calculation).
- For each month, calculate (Return − MAR/12). MAR is annual; divide by 12 for monthly.
- Replace any positive values with zero: =IF(value>0, 0, value).
- Square each downside deviation: =A2^2.
- Sum the squared downside deviations, divide by n (or n−1 for sample), take square root.
- Multiply by √12 to annualise.
- Sortino = (annual_return − annual_MAR) / annualised_downside_deviation.
Sources and methodology
Sortino ratio developed by Frank Sortino (1991). Standard formulation follows Sortino-Price (1994). MAR choice (3-month UK gilt yield) for consistency with Sharpe calculation. See the ETF Data Methodology for full data sources. The site methodology documents the broader review process.
Related metric pages
How UK Tax Drag holds itself to account
Every page is reviewed against the editorial standards, written from primary sources, sourced openly, and corrected publicly. No affiliate revenue. No sponsored content. No paid placements.