| Number | What it measures | The question it answers |
|---|---|---|
| Drawdown | The current fall from the most recent peak, as a percentage of that peak. | How far below my high-water mark am I right now? |
| Maximum drawdown | The deepest peak-to-trough fall anywhere in the period. | What is the worst this has ever been? |
| Drawdown duration | The time from the peak to the bottom of the fall. | How quickly did it go wrong? |
| Time under water | The time from the peak until a new peak is finally set. | How long would I have had to hold on? Usually the number people find hardest to accept. |
How to calculate maximum drawdown
Build the equity curve
Compound the return series into a running account value. Start at any figure — 1, or 100 — since drawdown is expressed as a percentage and the starting value cancels out.
Track the running peak
For each point in time, record the highest equity value seen up to and including that point. This is the high-water mark, and it never goes down.
Compute the drawdown at every point
For each point, take (equity − running peak) ÷ running peak. The result is zero at every new high and negative everywhere else.
Take the minimum
The most negative value in that series is the maximum drawdown. Report it as a positive percentage by convention — a maximum drawdown of 32% means the worst fall was 32% below a prior high.
Record when it happened, and how long it lasted
The date and the recovery time carry information the single percentage does not. A 30% fall that recovered in four months and a 30% fall that took six years are not the same risk.
Why does recovering from a drawdown take more than the loss?
Because the gain is calculated on a smaller base than the loss was. Lose 20% of 100 and you have 80; to get back to 100 from 80 you need 25%, not 20%. The asymmetry compounds and it gets brutal quickly.
- A 10% drawdown needs an 11.1% gain to recover.
- A 20% drawdown needs 25%.
- A 33% drawdown needs 50%.
- A 50% drawdown needs 100% — you have to double from the bottom just to get back to where you started.
- An 80% drawdown needs 400%.
This is why drawdown is the risk measure most people actually feel, and why it is not interchangeable with volatility. Two strategies with identical standard deviations can have very different worst falls, and the one with the deeper hole has a mathematically harder climb out of it regardless of how it performed on average.
What is the difference between drawdown and a loss?
A loss is measured from what you paid. A drawdown is measured from the best the account has ever been, whether or not you ever took that money out.
That distinction has a practical edge to it. A strategy can be up 40% on the year and simultaneously in a 15% drawdown, because it was up 65% two months ago. Nothing was lost relative to the starting point, and it will still feel like losing — which is why drawdown predicts whether somebody abandons a strategy better than the return does.
What does maximum drawdown not tell you?
It is a single observation, and it is the most extreme one in the sample. That makes it the least statistically stable number on any performance report.
- It says nothing about how often drawdowns happen. One 40% fall and twelve 15% falls can produce the same maximum drawdown as a strategy that fell 40% once and was otherwise calm.
- It grows mechanically with sample length. Run any strategy for longer and its maximum drawdown can only stay the same or get worse — never better — so comparing a two-year backtest to a twenty-year one on this number is comparing sample sizes.
- It says nothing about the recovery. The drawdown that ends the strategy is not the deepest one, it is the one nobody waits out.
- It is not a bound. A maximum drawdown of 25% over ten years is not a promise that 25% is the worst that can happen; it is the worst that did happen, in that sample.
Looking at the full distribution of drawdowns — how many, how deep, how long under water — carries far more information than the single worst figure, and almost nobody reports it.
Why is a backtest’s maximum drawdown usually optimistic?
Three reasons, and the third is the one that changes how you should read the number.
- Costs. Spread, slippage and financing come out of the equity curve continuously, and they deepen every fall. A drawdown computed before costs is not the drawdown you would have experienced.
- The sample. The backtest covers the regimes that happened to be in the data. A strategy that never traded through a liquidity crisis has a maximum drawdown that has never been tested against one.
- The search. If you tried many variants and kept the one with the best profile, the reported maximum drawdown is not an estimate of the worst case — it is the smallest worst case the search was able to find. Selection works on drawdown exactly as it works on the Sharpe ratio, and it is discussed far less often.
That third point is the one to carry away. The honest version of the question is not "what was the maximum drawdown" but "how many strategies did I look at before I found one whose maximum drawdown I liked".
Backtested and hypothetical results are not indicative of future performance. Nothing here is investment advice.
Common questions
What is a good maximum drawdown?
There is no threshold worth quoting, because the figure only means something next to the return that came with it and the length of the sample it was measured over. What is safe to say is that the number people believe they can tolerate is reliably larger than the one they actually tolerate when it happens.
How much do you need to gain to recover from a 50% drawdown?
100%. You have to double from the bottom to get back to the previous high, because the gain is computed on the reduced balance. This asymmetry is the main reason deep drawdowns matter more than their percentage suggests.
What is time under water?
The time from a peak until a new peak is finally set — not just the fall, but the whole round trip. It is often the more useful number, because it measures how long you would have had to keep holding a strategy that was visibly not working.
Does maximum drawdown depend on how long the backtest is?
Yes, mechanically. Maximum drawdown is the worst value in a sample, so extending the sample can only leave it unchanged or make it worse. A short backtest with a shallow maximum drawdown may simply not have run long enough to find a deep one.
Is drawdown the same as volatility?
No. Volatility measures how much returns vary around their average; drawdown measures how far the account fell below its own high. They are related but not interchangeable, and drawdown captures the path — the order the returns arrived in — which volatility ignores entirely.