The maximum drawdown is the largest peak-to-trough fall an account has had. It is a single event, and it depends on which trades happened to arrive together. The same strategy, run over a different period or with its trades in a different order, would have produced a different maximum. Treating that one number as the worst case is a mistake in both directions: it may have been a lucky run, or an unlucky one.
What the distribution shows
A strategy produces many drawdowns, not one. Over three years it might have had thirty falls from a peak: most of a few per cent, several of around ten, and one of twenty. The typical drawdown, the one that happens every few months, tells you what living with the strategy feels like. The maximum tells you the worst it has shown so far. Both are needed, and the worst it has shown so far is not the worst it can do.
Estimating what the future holds
One reasonable way to estimate a range is to take the strategy's actual trades and reshuffle their order many times, recording the maximum drawdown each time. Across thousands of shuffles, the drawdowns spread out, and the point below which 95 per cent of them fall is a more careful figure than the single historical maximum. Some verified-record services publish this as DD95. If a strategy's historical maximum is 15 per cent and its DD95 is 24 per cent, plan for 24, and accept that even that can be exceeded.
Length matters as much as depth
A record covering one calm year may show a maximum drawdown of 6 per cent. The same strategy across a year that included a market shock might show 18. The value of a maximum drawdown figure rises with the length and variety of the period it came from, and a figure from a short record should be treated as a lower bound.
Two EAs each show a maximum drawdown of 12 per cent. The first ran for eight months through a steady market. The second ran for four years including a crash, a rate-rise cycle and two flat periods. The first figure is one sample from one weather pattern. The second has been tested against most of what markets do. The number is the same. The information in it is not.
Read the maximum drawdown on an EA's record together with the length of the record, the number of separate drawdowns, and the typical size of them. If a DD95 or similar estimate is published, use that as the planning figure. Then choose the risk setting on the assumption that the next drawdown will be somewhat worse than the worst so far, because over a long enough period it will be.