Drawdown

A drawdown measures the worst of the fall rather than the size of a typical wobble. Maximum drawdown, the version usually quoted, is the largest peak to trough decline over a period. Unlike volatility, it depends on the order returns arrive in: shuffle the same set of monthly returns into a different sequence and the maximum drawdown changes, which is why it tracks what an investor actually lived through.

Two features make it worth quoting. Recovery is asymmetric, because the gain needed to get back to the peak is $1/(1-d) - 1$, where $d$ is the drawdown as a decimal. A 20 percent fall needs a 25 percent gain to undo it, a 50 percent fall needs 100 percent, and an 80 percent fall needs 400 percent. The second feature is time. A fall that takes six years to recover costs an investor something that a fall recovering in six months does not, so the length of a drawdown is reported alongside its depth.

The mistake is reading a maximum drawdown out of a backtest as a worst case. It is the worst that happened in one particular sample, not a limit, and the next one can be deeper. A longer track record also tends to show a larger maximum drawdown simply because it had more chances to find one, so a three year figure and a thirty year figure are not a like-for-like comparison. The measurement also always runs from the previous peak rather than from what you paid, so someone who bought part way up experiences a smaller fall than the headline number. For the spread measure that ignores the path entirely, see standard deviation of returns.