Monte Carlo Drawdown: Estimating Worst-Case Losses

The problem with a single equity curve

A backtest produces one equity curve, one sequence of wins and losses in the exact order they happened. But that order was partly luck. If your three worst losing trades had happened to land back-to-back instead of spread out, your maximum drawdown would have been far deeper, even though the set of trades was identical. Judging worst-case risk from a single historical ordering is like judging how dangerous a road is from one drive where you happened to hit every green light. Monte Carlo simulation exists to expose the danger your one lucky drive concealed.

How Monte Carlo drawdown analysis works

The method is conceptually simple. Take the set of returns or trade results your strategy produced. Then generate many alternative sequences, typically thousands, by resampling them, either by shuffling their order or by drawing from them at random with replacement. Each alternative sequence is one plausible history that the same strategy could have produced. For each one, measure the maximum drawdown, the deepest peak-to-trough decline. After thousands of runs, you have a distribution of possible worst cases rather than a single number.

That distribution is the payoff. Instead of saying the worst drawdown was 18 percent, you can say something far more useful: across thousands of plausible orderings, the drawdown exceeded 25 percent in a meaningful fraction of them, and in the most adverse runs it reached well beyond that. A conceptual example: a strategy whose actual historical drawdown was 15 percent might show, under resampling, that a 30 percent drawdown is entirely within the range of normal bad luck. If you sized your risk assuming 15 percent was the worst case, you were under-prepared.

Why it matters for survival

Risk of ruin is about the path, not the average. Many traders are forced out not because their strategy lacks an edge, but because a drawdown deeper than they prepared for arrives, breaches their tolerance or their margin, and ends the experiment. Monte Carlo drawdown analysis lets you rehearse those deeper drawdowns in simulation and ask an honest question: if this happened, could I withstand it without abandoning the strategy or being liquidated? Sizing your position so you can survive the adverse end of the distribution, not just the comfortable historical case, is one of the most practical uses of the technique.

Important limitations

Monte Carlo is powerful but not magic, and its limitations are easy to forget. First, it can only reshuffle the trades you already have; it cannot invent a kind of loss your strategy never experienced. If your history never contained a true crash, resampling that history will not conjure one, and your worst case will still be understated. Second, plain resampling typically assumes trades are independent, but real returns often cluster, with volatility arriving in bursts, so naive shuffling can actually understate tail risk. Third, the quality of the output depends entirely on the quality and length of the input trades; a short or biased trade history produces a confident-looking distribution built on thin evidence.

Common mistakes

The biggest mistake is treating the Monte Carlo worst case as a hard ceiling. It is an estimate of plausible bad luck given your data, not a guarantee that nothing worse can happen; the real world can always exceed your sample. Another error is running too few simulations to see the tail, or reporting only the median outcome and ignoring the painful percentiles that actually matter for survival. Used honestly, Monte Carlo drawdown analysis turns risk from a single comforting number into a sobering range, and it is the range, not the average, that determines whether you are still trading next year.

TRION treats drawdown as a range, not a single flattering number: resampling runs on real stored trade data with fixed seeds so they are reproducible, and the adverse percentiles are shown honestly rather than buried under a median.

Paper-only, simulation-only. No broker, no real orders, no promise of profit. Risk protects, humans decide.