A backtest hands you a single equity curve: these trades, in this order, produced this result. But the order was partly luck. The same strategy, started a month later or catching signals in a different sequence, would have produced a different curve — sometimes dramatically different. Monte Carlo simulation is the standard technique for seeing past that single draw of history.
The mechanics
Take the trade list from a backtest — say 120 trades with their individual percentage outcomes. Now reshuffle their order randomly and replay the sequence, compounding as you go. Do this five hundred or a thousand times. Each replay is an alternate history built from the same underlying performance; collectively they form a distribution of outcomes instead of a single number.
From the distribution you extract the statistics that actually govern survival: the median final equity (what 'typical' looks like once sequence luck is averaged out), the 5th-percentile outcome (a realistic bad case), the probability that maximum drawdown exceeds thresholds like 20% or 30%, and a risk-of-ruin estimate — the fraction of alternate histories in which the account fell below a survival floor.
What the results change
The most common discovery is that the backtest's drawdown flattered the strategy. A system showing a 12% historical max drawdown will often show a median resampled drawdown of 18% with a meaningful tail beyond 30% — the original sequence just happened to space its losers politely. That single insight should immediately inform position sizing: if the 95th-percentile drawdown at current size would break you (financially or psychologically), the size is wrong, whatever the average return says.
Monte Carlo results also separate luck from structure across strategies. Two systems with identical backtest returns can carry wildly different tail risks, and only the distribution view reveals it.
Limits, honestly stated
Simple resampling assumes trades are independent, which is not quite true — markets have regimes, and losses cluster. Treat Monte Carlo outputs as a lower bound on how bad things can get, not an upper one. Used that way, it remains one of the highest-value habits in strategy research: it converts 'my backtest made money' into 'here is the range of futures my evidence actually supports'. Educational tooling only — simulations model the past's trades and cannot guarantee any future outcome.