What Is the Calmar Ratio, and How Is It Different From Sharpe and Sortino?
The Calmar ratio divides a track record's compound annual growth rate (CAGR) by the absolute value of its worst peak-to-trough decline over a trailing window — commonly the last 36 months — instead of dividing by a measure of the full spread of returns the way the Sharpe and Sortino ratios do. That one substitution means Calmar doesn't ask "how much did returns bounce around, on average" at all; it asks one narrower question — "how much did the worst stretch cost, relative to what the strategy earned getting there."
The rest of this page covers how the ratio is actually calculated, why its single-worst-stretch design appeals to a different kind of trader than Sharpe or Sortino do, what window length changes about the number, and what it still doesn't tell you on its own.
How is the Calmar ratio calculated?
Take the compound annual growth rate over a window, divide it by the absolute value of the maximum drawdown over that same window.
- Pick a window — the original convention (built for CTA/managed-futures reporting in the late 1980s) uses a trailing 36 months, though some reporting uses the strategy's full history instead.
- Compute the CAGR for that window: the single annualized growth rate that would take the starting equity to the ending equity over that many years.
- Find the maximum drawdown over the same window — the single worst peak-to-trough decline, not an average of several declines. See what maximum drawdown is and why it matters more than win rate for exactly how that number gets found.
- Take the absolute value of that drawdown (it's naturally negative) and divide the CAGR by it.
A Calmar ratio of 1.0 means the window's annual growth rate exactly matched the size of its worst decline. A ratio above 1.0 means the strategy earned more per year than its single worst stretch cost; below 1.0 means the worst stretch was larger than a year's growth.
How is this different from Sharpe and Sortino?
All three ratios divide a return figure by some measure of risk — the difference is entirely in what that risk measure counts.
| Ratio | Risk measure (the denominator) | What it's really asking |
|---|---|---|
| Sharpe | Standard deviation of every period's return, up and down | How much did returns bounce around on average, in either direction? |
| Sortino | Downside deviation — the spread of only the returns below a chosen minimum | How much did returns bounce around on average, counting only the losing side? |
| Calmar | The single worst peak-to-trough decline over the window | How bad did the single worst stretch get? |
Sharpe and Sortino are both built from a distribution — every period in the window contributes to the denominator, averaged or partially averaged together. Calmar throws almost all of that information away and anchors on exactly one event: the deepest hole the equity curve ever had to climb out of. See what the Sharpe ratio actually measures and how the Sortino ratio differs from Sharpe for the full account of what those two keep that Calmar discards.
That design has a direct consequence: two strategies with identical Sharpe ratios can have very different Calmar ratios, because Sharpe can't see that one of them concentrated its bounce into a single ugly stretch while the other spread the same total bounce evenly across many smaller ones. Calmar is built specifically to catch the first case.
Why would a trader prefer the single-worst-stretch view?
Because for a lot of traders, the thing that actually ends a strategy isn't the average bounce — it's one stretch bad enough to break conviction, trigger a forced liquidation, or simply be more than the trader can tolerate watching happen to their account. A ratio built from an average doesn't answer "could I have sat through the worst thing that happened" — it answers a different, softer question.
This is also why Calmar has a long history in managed-futures and CTA reporting specifically: those are products sold partly on "here's the worst drawdown a real investor would have had to live through," and a ratio that puts that exact number in the denominator matches how the product is actually being evaluated by an allocator deciding whether they could stay in it.
The tradeoff is that Calmar is a one-event measurement. A strategy can have an excellent Calmar ratio and still have had a rough, choppy history full of smaller drawdowns that never individually became the single worst one — Calmar is silent on all of that, by design.
Why does the window length matter so much?
A short window hasn't had time to show the strategy's real worst case yet, and a window that happens to exclude the strategy's actual worst historical stretch will report a Calmar ratio that looks better than the full history supports.
- A trailing 36-month Calmar ratio rolling forward will drop sharply the moment a new worst drawdown enters the window, and it will rise again once that drawdown ages out the back of the window — even if nothing about the strategy's underlying risk has changed. The number moves with the window, not just with the strategy.
- A full-history Calmar ratio is more stable but can go stale for a strategy that has genuinely changed its risk profile since the drawdown that still anchors the denominator.
- Any window shorter than the strategy's slowest-forming cycle of bad luck is the same trap sample size sets for a win rate — not enough history has happened yet for the worst case to have shown up at all.
Quoting a Calmar ratio without stating the window it was computed over is close to meaningless — the same number can describe two very different periods in a strategy's life depending only on which months were included.
What does the Calmar ratio still not tell you?
- It says nothing about how often smaller drawdowns happened, only about the single worst one. A strategy with one deep hole and otherwise smooth results, and a strategy with constant shallow chop that never produced one standout bad stretch, can post similar Calmar ratios for very different reasons.
- A short or lucky window can simply be missing the real worst case. See how many calls before a track record means anything for the same caution applied generally, and how to spot a cherry-picked track record for the version of this specific to a window chosen after the fact to look flattering.
- A backtested Calmar ratio inherits the backtest's own limits. A rule set searched or tuned until it happened to dodge the market's worst historical stretches reports a shallower maximum drawdown — and a better Calmar ratio — than the strategy's real forward risk is likely to produce. See what curve-fitting is and how it fakes a backtested track record.
- There's no universal "good" cutoff. A ratio above 1.0 is commonly treated as a reasonable baseline in managed-futures reporting, but that convention doesn't transfer automatically to every strategy type or every window length.
- It doesn't distinguish a real account from a simulation. Percentages and ratios read the same whether they came from a live record or a backtest; only the source, stated honestly, tells you which one you're looking at.
Questions traders ask
Is a higher Calmar ratio always better than a higher Sharpe ratio?
Neither is "better" in general — they measure different things. Calmar is the more direct answer when the question is specifically "how bad was the single worst stretch relative to what the strategy earned." Sharpe (or Sortino) is the more complete answer when the question is about the whole distribution of bounce across the window, not just its single worst event.
Can the Calmar ratio be negative?
Yes. If the CAGR over the window is negative — the strategy lost money overall across that period — the numerator goes negative and the ratio goes negative with it, since the denominator (the absolute value of the drawdown) is always positive.
Why 36 months specifically?
There's nothing mathematically special about three years — it's a convention carried over from the managed-futures reporting practice the ratio originated in, chosen as a window long enough to likely contain at least one real adverse stretch without going so long that older, possibly obsolete performance dominates the number. Many providers report Calmar over the full track record instead, and the two conventions aren't directly comparable without knowing which was used.
Does a strategy need a long track record before its Calmar ratio means anything?
Yes, for the same reason any drawdown-based measurement needs enough history to have plausibly seen its real worst case. A new strategy's Calmar ratio reflects whatever drawdown has happened to occur so far — which, for a strategy still early in its life, usually isn't its eventual worst one.
The Calmar ratio trades the full-distribution view Sharpe and Sortino take for a single, specific worst-case anchor — which makes it the more direct read for a trader whose real question is "could I have sat through the worst thing that happened here," and a narrower one for every other question a track record raises. None of the three ratios replaces checking how long the record actually is, whether it's live or simulated, and whether the window it was computed over was chosen honestly. How any of this gets read in practice, rather than just defined, is on the methodology page.
Keep reading
Get them free