The Monte Carlo page described how reshuffling a strategy's trades produces a distribution of maximum drawdowns. This page is about reading that distribution once it exists, because the figures in it answer questions that a single historical maximum cannot.
The figures
The median maximum drawdown is the middle of the range: half of the simulated histories were worse. DD95 is the level that 95 per cent of them stayed under, and it is the usual planning figure. DD99 is the one-in-a-hundred figure, and the gap between DD95 and DD99 shows how heavy the tail is. Alongside these, the distribution of time under water, the number of months from a peak to the next new peak, tells you how long a drawdown is likely to last, which decides whether it is bearable.
Reading them together
A strategy with a DD95 of 18 per cent and a median recovery of three months is a different proposition from one with a DD95 of 18 and a median recovery of eleven months. The second requires patience the first does not, and most owners do not have it. The tail matters too: a DD99 of 40 against a DD95 of 18 means the strategy has an ugly minority of outcomes, usually from a few very large losing trades, and the position sizing should look at those trades.
Scaling with risk
Drawdown scales roughly with risk per trade. A DD95 of 18 per cent at one per cent risk is about 34 at two per cent and about 9 at half a per cent. The distribution is computed once and read at whichever risk setting is under consideration, which makes it the tool for choosing the setting: find the risk at which DD95 sits at the drawdown you would actually sit through, and use that.
A strategy's distribution at one per cent risk shows a median maximum drawdown of 12 per cent, a DD95 of 19, a DD99 of 26, and a median time under water of four months with a 95th percentile of ten. An owner who knows they would stop at 15 per cent down should not run this at one per cent, because there is a one-in-four chance or so of seeing more than 15. At 0.7 per cent, DD95 is about 13, and the same owner has a strategy they can hold.
Where an EA's record offers a drawdown distribution, use it in this order: read DD95 and the recovery time at the risk setting you intend, compare DD95 with the drawdown you would tolerate, and adjust the setting until they match. Where only a historical maximum is offered, treat it as roughly the median of a distribution you cannot see and plan for something around half again as deep. And remember the serial correlation page: the reshuffled figures are a floor, not a ceiling.