What Is the Calmar Ratio?

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-14 · Updated 2026-10-03

Most performance ratios treat all variation as risk. Calmar treats only the holes as risk, which is much closer to what anybody actually experiences holding a position.

How it works

Annual return on top, maximum drawdown underneath. A strategy returning 20% a year with a worst drawdown of 10% has a Calmar of 2.0.

Higher is better, and the improvement can come from either side. Raise the return or reduce the worst hole, and both are genuine improvements rather than accounting choices.

A big return earned through a brutal drawdown scores poorly, which is the whole design. Two strategies returning 30% are not equivalent if one did it smoothly and the other halved on the way.

Why the worst case and not an average

Because you only have to be destroyed once. An average drawdown of 8% is no comfort if one of them was 60% and the account did not survive it.

That is also the criticism. A single event — possibly one unrepeatable market day — sets the denominator for the entire history, so the figure is hostage to one observation.

Both things are true at once. It is the right question and a fragile measurement, and anyone quoting it should say over what period.

A worked example

Take a hypothetical $10,000 account that ends the year at $11,200, and whose worst dip took it from $10,500 down to $9,450. The return is 12% and the deepest drawdown is 10%.

Calmar on those figures is 1.2 — return slightly larger than the worst hole.

Now consider what a short sample would have shown. Pick a window that happened to miss that 10% drawdown and the denominator collapses — the same strategy reports a Calmar several times higher without anything about it changing.

That is the flattery problem, and it is not subtle. A two-year track record has had two years to encounter its worst case. A twenty-year one has had twenty.

So a high Calmar over a short history is close to meaningless, and the figure should always be read alongside how long it was measured over.

What it does not capture

How long the drawdown lasted. A 20% hole recovered in a month and one that took four years produce the same denominator, and they are completely different to live through. The S&P 500 once took 7.5 years, from January 1973 to July 1980, to close above an old record — a duration Calmar does not see.

How many drawdowns there were. One deep hole and fifteen shallow ones can score identically.

And what happens next. Every figure here is historical, and the worst drawdown in the record is simply the worst one so far.

The original data

Across the S&P 500’s daily closes from 1950 to 2026, the deepest fall was 56.8%, from 1,565.15 on 9 October 2007 to 676.53 on 9 March 2009; 93% of trading days ended below an earlier record; and the longest wait for a new one ran 1,897 trading days, from 1973 to 1980.

That 93% figure is the context Calmar compresses into one number. Being underwater was the normal state, not the exception — so a ratio built on the single deepest point is summarizing a condition that was almost permanent.

And costs come out before the return does — the spread and commission on every round trip lower the numerator, so any Calmar calculated gross overstates what an investor would have kept.

How it compares to the alternatives

Sharpe divides excess return by total volatility. That counts upside movement as risk, so a strategy that occasionally leaps upward is penalized for it — which does not match how anybody experiences a large gain.

Sortino fixes half of that by counting only downside deviation, and it still measures deviation rather than depth. A strategy can have modest downside deviation and still have spent two years in a hole.

Calmar is the one that looks at the hole itself. Not how variable the path was, but how far down it went at its worst — which is the number that decides whether a position gets abandoned, and whether the capital was still there to continue.

The trade-off is statistical. Sharpe and Sortino use every observation, so they are stable and use the full sample. Calmar uses one observation, so it is unstable and asks the more relevant question. Reading two of them together is more informative than arguing about which is correct.

How this connects

The ulcer index answers the question Calmar ignores, measuring the depth and the duration of every decline rather than only the single worst one.

Its main alternatives scale return by standard deviation, which counts a sudden rise as risk in exactly the way a drawdown ratio refuses to.

It grew up in the managed futures and hedge fund world, where investors choosing between managers cared most about how deep the worst hole had been.

A backtest produces the most flattering Calmar of all, because the rules were chosen with the worst historical decline already in view.

Overfitting inflates it directly. A rule tuned to sidestep one historical crash reports a small denominator that will not survive live trading.

Walk forward testing is the honest check, measuring the ratio on data the rules were never fitted to.

A prop firm applies the same idea as a hard limit: the drawdown rule ends the account whatever the return was.

Expectancy sits underneath the return figure, and a positive one says nothing about how deep the holes between the winners will be.

And position sizing moves both halves at once: smaller size shrinks the return and the worst decline together, which is why the ratio changes less than either figure does.

When it fails

The characteristic failure is comparing Calmar ratios across different track-record lengths. A fund with three years of history and a Calmar of 4 looks superior to one with fifteen years and a Calmar of 1.5, and the comparison is almost meaningless — the longer record has simply had more opportunity to meet its worst case. Ranking managers this way systematically favors the ones who have not yet been tested, which is the opposite of what the ratio was built to do.

A second failure is using it on a strategy with rare, large losses. Selling options can post an excellent Calmar for years and the denominator is simply waiting.

A third is quoting it gross of fees and costs, which inflates the numerator.

A fourth is treating the worst drawdown as a limit. It is the largest observed, not the largest possible.

The question: Calmar of four over two years. Meaningful?

And a fifth is ignoring duration. Depth and length are different kinds of pain, and only one of them is in the formula.

One practical note on where it came from. Terry Young published it in 1991 for managed futures funds, where drawdowns are the thing investors actually redeem on. The convention there is a 36-month window, which is short enough that the flattery problem above is built into the standard usage rather than being an abuse of it.

Related

Risk of ruin covers what a deep drawdown does to recovery arithmetic. Absolute return covers the numerator. And volatility covers the variation Calmar deliberately declines to count.

This page is educational, not financial advice. Test every idea on your own charts before risking money.