A backtest gives you one sequence of trades. The drawdown you see is the worst drop in that particular order. If the same trades had arrived in a different order, with several losers in a row, the drop could have been much deeper. Monte Carlo trade shuffling estimates how much.
The idea
Take the list of closed trades and their returns. Shuffle the order randomly, rebuild the equity curve and measure its maximum drawdown. Repeat many times. You now have a distribution of drawdowns that the same set of trades could have produced.
How Backte.st does it
The Trade-order shuffle check reshuffles the trade order 200 times and reports the 95th-percentile drawdown: a level that only 5% of the shuffled sequences exceeded. It uses a fixed random seed, so the same report always gives the same number.
Why it is useful
- Sizing. If the shuffled drawdown is far deeper than the one in the backtest, you may want to risk less per trade.
- Expectations. Knowing that a 30% drop is plausible makes it easier to stick with a strategy through a bad stretch, or to decide in advance that you can’t.
- Comparisons. Two strategies with similar returns can have very different shuffled drawdowns.
Limits to keep in mind
- It ignores losses within trades. The shuffle works on closed-trade returns, so drawdowns while a position is open aren’t included.
- It assumes trades are independent. In reality, losing trades often cluster during certain market regimes, which shuffling breaks apart.
- It can’t invent new kinds of trades. If history never included a crash, neither will the shuffle.
- It needs enough trades. With very few, the distribution is meaningless, and the check reports too few trades.
Using it with the other checks
The shuffle answers “how bad could it get with these trades?”. The other checks ask whether the trades themselves are trustworthy: whether they hold up out of sample, after higher costs and without the best few winners.
Open the robustness checks for any run in the lab to see the shuffled drawdown next to the actual one.