Drawdown, recovery and equity curve shapes
Lesson 17 · about 10 min
Return tells you what a strategy made. Drawdown tells you what it cost to hold on long enough to make it. Of the two, drawdown is the one that decides whether you will still be trading the system when the return arrives, so it deserves at least as much attention.
Defining drawdown
Drawdown at any point is the distance from the running peak of equity to the current equity. Maximum drawdown is the largest such distance over the whole test. Measure it in R (so it does not depend on the sizing you happened to use) and also in percent of account (so it reflects compounding at the sizing you plan to use).
| Date | Equity (R) | Running peak | Drawdown |
|---|---|---|---|
| Jan | 10.0 | 10.0 | 0.0 |
| Feb | 14.5 | 14.5 | 0.0 |
| Mar | 11.2 | 14.5 | 3.3 |
| Apr | 8.7 | 14.5 | 5.8 |
| May | 7.1 | 14.5 | 7.4 |
| Jun | 9.9 | 14.5 | 4.6 |
| Jul | 13.0 | 14.5 | 1.5 |
| Aug | 15.8 | 15.8 | 0.0 |
Maximum drawdown in this stretch is 7.4R, and the time from the February peak to the August recovery is six months. Both numbers matter. A 7.4R drawdown at 1% risk per trade is a little under 7.5% of the account; at 2% it is close to 14%.
Recovery and time underwater
Time underwater is the number of bars (or days, or months) between a peak and the next time equity exceeds it. Most strategies spend more than half their life in some drawdown; that is normal. What you need to know is the longest one, because that is the period you will have to sit through while wondering whether the system is broken.
A useful table to build from the equity curve:
| Drawdown episode | Depth (R) | Depth (%) at 1% risk | Duration to trough | Duration to recovery |
|---|---|---|---|---|
| 1 | 7.4 | 7.2% | 3 months | 6 months |
| 2 | 5.1 | 5.0% | 5 weeks | 4 months |
| 3 | 11.2 | 10.7% | 4 months | 11 months |
| 4 | 4.0 | 3.9% | 2 weeks | 7 weeks |
Episode 3 is the one to think about: eleven months underwater, four of them getting worse. Would you have kept trading it? If not, the backtest's return is not available to you, because you would have stopped before it arrived.
The return-to-drawdown ratio
Annual return divided by maximum drawdown (sometimes called MAR or Calmar when computed over a specific window) is a plain-language alternative to Sharpe. A strategy making 15R a year with an 11R max drawdown has a ratio of about 1.4. Values above 1 are respectable for a single system over a multi-year test; values above 3 on a backtest should be checked for the usual errors. As with every ratio, it is unstable on short samples, because the max drawdown in a short test is not the max drawdown the strategy is capable of.
Key idea: Maximum drawdown in a backtest is the worst the strategy did in that sample, not the worst it can do. Expect the live maximum to be deeper. Plan sizing from the assumption that it will be, which Module 8 covers.
Equity curve shapes and what they mean
The shape of the curve tells you where the return comes from and what could take it away.
| Shape | Description | Likely source | Risk to watch |
|---|---|---|---|
| Steady upward slope | Small, consistent gains | High-frequency edge or mean reversion | Fat left tail not yet in sample |
| Staircase | Long flats, then jumps | Trend following; a few large winners | Dependence on the tail; long flat spells |
| One regime | Great for two years, flat before and after | A regime-specific edge | Regime ended; test on other periods |
| Hockey stick at the end | Flat, then a spike in the final months | Recent trend or one huge trade | Nearly all profit from a short window |
| Smooth, then cliff | Consistent gains, then a large loss | Short volatility, selling tails | The cliff will recur |
| Sawtooth around zero | No net direction | No edge | Do not trade it |
The "one regime" shape is the most common way a backtest fools a careful trader. Everything checks out, costs are honest, no leaks, and the whole return came from 2020 to 2021. Slice the curve by year and look at each year's contribution.
| Year | Net R | Share of total |
|---|---|---|
| 2018 | +2.1 | 6% |
| 2019 | +1.4 | 4% |
| 2020 | +14.8 | 44% |
| 2021 | +12.5 | 37% |
| 2022 | −1.9 | −6% |
| 2023 | +4.8 | 14% |
Eighty-one percent of the return came from two years. The honest description of this strategy is "works well in strong trends and is roughly flat otherwise", which is a legitimate thing to trade if you know it, and a trap if you expected the average.
Underwater chart
Plot drawdown over time as a chart that sits at zero and dips below it. It is the most useful single picture of a strategy's character, better than the equity curve, because it shows how often, how deep, and how long, on one axis.
Try it: From your backtest's equity series, compute the drawdown series and build the episode table above (depth in R and percent, duration to trough, duration to recovery) for every drawdown deeper than 3R. Then slice net R by year. Write down the longest time underwater and the share of return from the best two years.
Recap
- Drawdown is distance from the running peak; measure max drawdown in R and in percent.
- Time underwater, especially the longest episode, decides whether you would have survived to see the return.
- Return ÷ max drawdown is a plain alternative to Sharpe; above 1 is respectable, above 3 is suspicious.
- Backtest max drawdown is a sample statistic; the live one will likely be deeper.
- Read the curve shape and slice by year; a return concentrated in one regime is a different strategy from the one the average describes.