If you reshuffle the trade order of a backtest 10,000 times, the drawdown distribution is brutal — the median max drawdown is often 1.5-2x the single historical sequence you actually tested.
Question for people who do this: which percentile do you use to set position size? 90th? 95th? 99th?
95th percentile — and then I halve the size anyway, because live slippage adds another layer the simulation doesn't see.
Rule of thumb: if the 95th percentile drawdown is 40%, I want the account to survive 40% without a margin call and without me panicking and intervening. The second constraint is usually the binding one.
Same, 95th. Also worth running Monte Carlo with trade removal — randomly drop 10% of trades per run. It tests whether your edge depends on a handful of outlier winners. Many backtests quietly do.