A Monte Carlo simulation, in a trading context, is a statistical technique that reshuffles the order of a strategy's historical wins and losses thousands of times to estimate the range of drawdowns and loss streaks the strategy could plausibly produce, rather than relying on the single sequence that happened to occur in the actual backtest.
The problem with trusting one backtest
A backtest showing 60 wins and 40 losses over the past year tells you the net result of one specific ordering of those trades. If eight of those forty losses happened to land consecutively near the start, the equity curve would look far more painful than if they were spread evenly throughout the year, even though the underlying edge of the strategy is identical in both cases. Since the real sequence of future trades is unknowable in advance, judging a strategy only by the one historical path it actually took understates how bad things could get.
What running the simulation actually involves
The process starts with a verified list of past trades, including entry, exit, and result in R-multiples. A bootstrap algorithm then randomly draws from that same pool of outcomes, with replacement, to build thousands of alternate equity curves, commonly ten thousand runs. Across all of those runs, the simulation produces a full probability distribution for maximum drawdown, time to recover from a drawdown, and the stability of the strategy's Sharpe ratio, rather than a single number.
Turning the output into a sizing decision
If the original backtest showed an 8% maximum drawdown but the Monte Carlo distribution reveals that 5% of simulated paths experience a drawdown of 22% or worse, position sizing needs to be built around surviving that 22% scenario, not the friendlier 8% figure from the single historical path. A similar logic applies to risk-of-ruin estimates: if the simulation shows more than a small fraction of a percent chance of the account being wiped out entirely, cutting risk per trade, for example from 2% down to 1%, meaningfully changes that probability even though it might feel like giving up upside in the short run. Setting a portfolio-level stop-loss based on the 99th percentile drawdown from the simulation, rather than the best-case historical number, is what actually keeps a strategy tradable through its worst realistic stretch.