Calmar divides return by the worst drawdown — a brutally honest survivability score. But a return and a drawdown over what period? The textbook answer is a trailing 36 months. Most calculators quietly pick something else. Here's the worked math showing how much that one setting moves the number, and why the bias only ever runs one direction.
A survivability score is only honest if you know its window.Opere en Bybit: gran liquidez, tarifas bajas y un bono de bienvenida de hasta $30 000.Abrir cuenta Bybit →Calmar prices return against the single worst peak-to-trough loss, not average volatility — it answers "how much annual return am I earning for the deepest hole I'd have to sit through?"
Calmar = annualized return (CAGR) ÷ maximum drawdown
The part almost nobody states out loud is CAGR and max drawdown over what time period? Terry W. Young's original 1991 definition (built for the California Managed Accounts Reports he was tracking — hence "Calmar") used a rolling trailing 36-month window for both halves of the ratio. Decades later, most public Calmar numbers you'll see — on a fund fact-sheet, a backtest report, a strategy leaderboard — use whatever window the data happens to cover, which is usually full history since inception, not the textbook 36 months. Same formula, different window, different number.
Take a strategy with 5 full years of annual returns and each year's worst intra-year drawdown. Year 1 included a sharp 55% crash; the last three years have been calm:
| Año | Devolver | Peor reducción | In trailing 36mo? |
|---|---|---|---|
| Year 1 (oldest) | +20% | 55% | No |
| Year 2 | +25% | 12% | No |
| Year 3 | +18% | 10% | Sí |
| Year 4 | +22% | 9% | Sí |
| Year 5 (most recent) | +20% | 8% | Sí |
Compounding all 5 years gives a full-history CAGR of (1.20×1.25×1.18×1.22×1.20)^(1/5)−1 = 21.0%. Compounding only the trailing 3 years gives (1.18×1.22×1.20)^(1/3)−1 = 20.0% — the CAGR barely moves between windows. The maximum drawdown is where the real swing happens:
| Window | CAGR | Reducción máxima | Calmar |
|---|---|---|---|
| Full history (5 years) | 21.0% | 55% | 0.38 |
| Trailing 36 months (official) | 20.0% | 10% | 2.00 |
Same strategy, same yearly returns, same drawdowns — Calmar reads 0.38 under full history y 2.00 under the textbook trailing 36-month window, a swing of more than 5x, purely because the old 55% crash sits outside the shorter window and the full-history number can never un-see it.
With Sortino, switching the MAR setting can push the ratio up or down depending on the shape of the return distribution. Calmar's window-length bias is more one-sided, for a simple structural reason: a longer window's maximum drawdown can never be smaller than the maximum drawdown of any shorter window it fully contains. Every peak-to-trough decline inside the trailing 3 years is also a valid candidate decline inside the full 5-year history — so the full-history max is always at least as large. Shrinking the window can only hold the drawdown steady or shrink it; it can never reveal a worse one that the longer window had missed.
That means, holding CAGR roughly constant, a shorter trailing window can only flatter the Calmar ratio or leave it unchanged — never make it worse. The one way a trailing-window Calmar still comes out lower is if the trailing period's own CAGR is meaningfully weaker than the full-history CAGR — the numerator shrinking by more than the denominator does. That can happen (a strategy that was hot early and has cooled off recently), but in most real track records the drawdown effect dominates, which is exactly why a strategy's trailing-36-month Calmar is so often the more flattering number to publish.
The textbook Calmar window needs 36 months of history to exist at all. A strategy with 14 months of live results can't compute a trailing-36-month Calmar — whatever a dashboard displays for it is necessarily a full-history number, even if it's labeled "Calmar (3yr)" next to older, genuinely-3-year peers. Comparing a young strategy's full-history Calmar against an established fund's trailing-36-month Calmar compares two different statistics wearing the same label, independent of which one actually performed better.
The Calmar ratio was originally defined by Terry W. Young in 1991 using a trailing 36-month (3-year) window for both the annualized return and the maximum drawdown. In practice, many calculators, dashboards and fund fact-sheets quietly swap in a full-history (since-inception) window instead, or a calendar-year window — none of which is the textbook definition, and each produces a different number from identical underlying data.
On a 5-year track record with a 55% drawdown in year one and a smooth last three years (18%, 22%, 20% returns, 10% worst drawdown), the full-history Calmar ratio is 0.38 (21.0% CAGR ÷ 55% max drawdown) while the trailing 36-month Calmar ratio is 2.00 (20.0% CAGR ÷ 10% max drawdown) — more than 5x higher. The CAGR barely moves between windows; the maximum drawdown is what swings, because the old crash simply falls outside the shorter window.
The maximum-drawdown component can only be equal to or smaller in a shorter trailing window than in the full history that contains it — a longer window can never have a smaller worst peak-to-trough decline than any sub-window inside it. So the drawdown half of the ratio is structurally biased toward flattering shorter windows. The one way trailing Calmar can still end up lower is if the trailing period's annualized return is disproportionately weaker than the full-history return — possible, but less common than the drawdown effect dominating.
Find out what window each one used — trailing 36 months, full history since inception, or calendar year-to-date — before treating the numbers as comparable. A fund advertising a 36-month Calmar right after a strong 3-year run looks structurally better than the same fund's since-inception number, not because performance changed, but because an older drawdown rolled out of the measurement window. Also check how young the track record is: a strategy under 36 months old cannot compute the textbook trailing window at all, so its reported Calmar is really a full-history number regardless of label.
⚠️ Solo educación, no asesoramiento financiero. Las operaciones con apalancamiento pueden hacer perder todo su depósito.