What Is Maximum Drawdown, and Why It Matters More Than Win Rate
Maximum drawdown is the largest percentage decline a track record ever suffered from a prior peak before making a new high — not the average dip, the single worst one. It matters more than win rate, and sometimes more than expectancy, because it measures the one thing both of those numbers skip: how deep the hole gets before the system proves it can climb back out, and how much that hole costs to climb out of.
That's the answer to the search query. The rest of this page covers how to calculate it, why a high-win-rate or positive-expectancy system can still carry a brutal drawdown, the asymmetric math that makes a big loss worse than it first looks, and what the number doesn't tell you on its own.
How is maximum drawdown calculated?
Track the running peak of an equity curve — the highest value it has reached so far — and measure every decline from that peak until a new peak is made. The largest of those declines, as a percentage, is the maximum drawdown.
- Start at the first equity value and call it the running peak.
- At every new point, if the value exceeds the running peak, update the peak.
- If it doesn't, compute the percentage decline from the peak to the current value.
- The largest decline recorded across the whole series is the maximum drawdown.
It's a single number about the worst stretch, not an average. A track record can have a mild, boring average and still carry one five-week stretch that erased a third of its equity — drawdown is built to surface exactly that stretch, which an average would smooth over.
Why can a system with a high win rate still have a deep drawdown?
Because win rate counts how often a system wins, and drawdown measures how badly its losses can cluster — two different questions with two different answers. A system built on win rate vs. expectancy's "System A" shape — frequent small wins, one rare large loss — can run a long, satisfying winning streak and then give back months of gains in the single stretch that lands on that large loss. Nothing about winning 7 of every 10 trades rules that out; a high win rate only describes how often the small wins show up, not how big the rare loss is sized to be.
The same logic runs the other direction. A system with a low win rate and a positive expectancy — "System B" in that same comparison — can carry a shallower drawdown than its win rate alone would suggest, if its losses are kept consistently small even while they're frequent. Expectancy and drawdown both look past the win rate, but they're still answering different questions: expectancy asks what a trade is worth on average, and drawdown asks how bad the worst stretch got along the way.
Why does a bigger loss need a disproportionately bigger gain to recover from?
Because the percentage gain needed to recover from a loss grows faster than the loss itself — the math is asymmetric, and it gets worse the deeper the drawdown goes.
| Drawdown | Gain needed to get back to even |
|---|---|
| −10% | +11.1% |
| −20% | +25% |
| −30% | +42.9% |
| −50% | +100% |
| −75% | +300% |
| −90% | +900% |
A 10% loss only needs an 11.1% gain to erase — close enough that it barely registers as asymmetric. A 50% loss needs a 100% gain, not a 50% gain, just to get back to where it started. Past that point the math turns punishing fast: a 75% loss needs the remaining capital to quadruple, and a 90% loss needs it to multiply by ten. This is why maximum drawdown is read as a risk number in its own right rather than folded into an average — the cost of a loss isn't linear, so the depth of the worst one matters disproportionately more than its size alone suggests.
Does a shallower drawdown always mean a better system?
Not by itself — drawdown has to be read next to what produced it, the same caution that applies to win rate and expectancy on their own.
- Sample size still applies. A short track record simply hasn't lived through enough market conditions to show its real worst case yet. The deepest drawdown on record is a floor on what's possible, never a ceiling — see how many calls it takes before a track record means anything for the same logic applied to win rate.
- A backtest's drawdown can be curve-fit like anything else in it. A rule set searched until it happened to dodge the worst historical stretch will report a smaller maximum drawdown than the strategy's real exposure to risk — see what curve-fitting is and how it fakes a backtested track record.
- A cherry-picked record can simply omit the drawdown. Showing only the winning stretch of an account, or starting the clock after the worst period already happened, produces a shallow-looking drawdown that never happened in the full history — one of the exact tells covered in how to spot a cherry-picked track record.
- Duration is a separate question from depth. A 20% drawdown that resolves in three weeks and a 20% drawdown that drags on for two years are very different experiences to live through, and the maximum-drawdown figure alone doesn't distinguish them.
What does maximum drawdown not tell you?
- It doesn't tell you if the drawdown is over. The number is backward-looking by definition — the worst decline so far — and a live system can always be in the process of carving out a new, deeper one right now.
- It doesn't describe the path. A straight, steady decline and a jagged, whipsawing one can produce the same maximum-drawdown number while feeling completely different to sit through.
- It doesn't include real trading costs unless they were already in the equity curve. Commissions, spread, and slippage all widen a loss in practice; a drawdown computed on raw price moves can understate what a live account would have actually gone through.
- It doesn't say anything about expectancy. A system can have a shallow maximum drawdown and still be a net loser, or a deep one and still be a net winner — drawdown measures risk endured, not profitability. Reading it alongside win rate vs. expectancy gives a fuller picture than either number alone.
Questions traders ask
Is maximum drawdown the same as volatility?
No. Volatility measures how much an equity curve fluctuates generally, in both directions; maximum drawdown measures one specific thing — the single worst decline from a peak. A curve can be highly volatile day to day while never suffering a large drawdown, if the ups and downs mostly offset each other before a new high is set.
How much drawdown is "too much"?
There's no universal number — it depends on what the capital is for and how long a decline of that size can be tolerated before a real decision (stopping, resizing, walking away) gets forced. The honest comparison is always the drawdown's depth and duration together against what the person holding it can actually sit through, not a single threshold that applies to everyone.
Can a system's future drawdown be bigger than anything in its history?
Yes. Every maximum-drawdown figure describes the worst decline observed so far in a finite history, not a ceiling the system is bound by. A longer track record, or a market condition the system hasn't yet lived through, can always produce something worse than the current record shows.
Does a small backtested drawdown mean a strategy is safe to trade live?
Not on its own. A backtested drawdown inherits every limitation the backtest has — sample size, whether it was curve-fit, whether real costs were included — before it says anything about what a live account would actually experience. The questions above about sample size and curve-fitting apply here exactly as they do to win rate and expectancy.
Where this fits the bigger picture: drawdown, expectancy, sample size, and clean (not cherry-picked or curve-fit) data are the legs a track record needs before any single number from it means much. The Sharpe ratio folds return and bounce into one number the same way drawdown folds decline into one number — reading the two together catches more than either alone. The one reported drawdown is also just one order of the same trades — Monte Carlo simulation reshuffles that order thousands of times to show how much worse the same trades could plausibly have looked. The drawdown a position size can actually produce is also the real-world cost side of the Kelly Criterion — which is exactly why full Kelly's theoretical growth rate is rarely traded straight. How the method behind this site's own graded calls holds up against that bar is on the methodology page.
Keep reading
Get them free