The Calmar Ratio: Why I Judge a Strategy by Its Worst Drawdown, Not Its Best Month

When I show someone a backtest, the first thing they look at is the profit number. I understand the instinct, I did the same when I started. But after building and rejecting enough strategies, the first number I look at now is a different one: maximum drawdown. And the ratio that puts it front and center is the Calmar ratio.

This post is the plain-language version of why. No hype, just the reasoning.

What is the Calmar ratio?

The Calmar ratio is a simple division:

Calmar = annualized return ÷ maximum drawdown

If a strategy returns 30% per year and its deepest peak-to-trough equity drop was 15%, its Calmar is 2.0. If another strategy also returns 30% but once sat in a 60% hole, its Calmar is 0.5.

Same headline return. Completely different strategies to actually live with. (These numbers are illustrative, not results from any KenKem system.)

The denominator, maximum drawdown, is the deepest valley your account equity fell into before making a new high. It answers a very concrete question: at the worst moment, how much of my capital had evaporated on paper?

Why not just use the Sharpe ratio?

The Sharpe ratio (return divided by volatility) is the most famous risk-adjusted metric, and it is genuinely useful. But it has a property that never sat right with me as an engineer: it penalizes volatility symmetrically. A strategy that lurches upward gets punished the same as one that lurches downward.

No trader has ever complained about upside volatility.

The Calmar ratio only cares about the thing that actually hurts: the deepest hole. That matches how risk is experienced in reality, by traders, by prop-firm risk desks, and by anyone allocating capital. Prop firms don't set Sharpe limits; they set hard drawdown limits. Breach the drawdown, and the account is gone regardless of how elegant the strategy was.

Why is drawdown the number that actually breaks traders?

Because almost nobody abandons a strategy over a statistics dip. They abandon it because they are 25% down, the recovery is nowhere in sight, and every fiber of their psychology is screaming to pull the plug, usually at the exact bottom.

I have felt this with real money in manual trading. It is precisely the failure mode I am trying to engineer out of my own trading with a systematic approach. Rules and automation can remove emotion from entries and exits, but drawdown is where emotion attacks the system itself: "turn it off, it's broken."

So when I evaluate a strategy, maximum drawdown isn't just a risk statistic. It is a proxy for a very practical question: will the human running this system actually survive running it? A strategy with a modest return and a shallow drawdown is often worth more than a spectacular backtest nobody could psychologically hold through.

What are the Calmar ratio's blind spots?

Honesty requires saying where the metric is weak:

Max drawdown is a single sample. It is one number from one historical path. Shuffle the same trades into a different order and the max drawdown changes, sometimes a lot. A strategy that showed a 12% drawdown historically might plausibly produce 20% on a different-but-equally-likely sequence of the same trades.

It says nothing about frequency. One deep drawdown five years ago and chronic 10% drawdowns every quarter can produce similar Calmar values but feel completely different to trade through.

It is period-sensitive. Calmar computed over a lucky two-year window flatters a strategy; the same strategy over a window containing a regime change can look terrible.

The engineering answer to the first problem is to stop treating max drawdown as one number: resample the trade sequence many times (Monte Carlo) and look at the distribution of drawdowns. "95% of resampled paths stayed above X% drawdown" is a far more honest statement than "the backtest drew down Y%."

There is also a direct link between the drawdown you can tolerate and how large you can size positions, that is the territory of the Kelly criterion, which deserves its own post.

How does this shape the KenKem workflow?

At KenKem we are building software for systematic Gold and Crypto trading, TradingView indicators and MT5 Expert Advisors, and drawdown-first evaluation is baked into how we test our own work. To be transparent: our systems do not yet have a live track record, and we will not publish performance claims until real forward results exist. What we can share is the method: real tick data, walk-forward testing, and judging every candidate by its drawdown behavior before its return.

The same philosophy shows up in the software itself: the user, not KenKem, configures risk per trade, maximum daily risk, and maximum drawdown limits, and the EA follows those settings. Capital preservation settings exist precisely because the drawdown, not the profit target, is what decides whether a trader is still in the game next year.

If you take one thing from this post: when someone shows you a return number, ask for the drawdown that bought it. The ratio of the two tells you more than either number alone.

Educational purpose only. Not financial advice. Backtests and historical examples do not guarantee future results, users remain responsible for their own trading decisions and risk settings.