Monte Carlo Simulation

Quick Answer

Monte Carlo simulation reshuffles or resamples a backtest's trades thousands of times to build a distribution of possible equity paths instead of trusting the one history delivered. It shows how bad drawdown could get: a system with 18% historical drawdown might show a 95th-percentile drawdown of 30%. It assumes trades are independent, which options strategies often violate.

Definition: Monte Carlo Simulation

Monte Carlo Simulation is a robustness technique that resamples or reorders a backtest's trades many times to produce a distribution of outcomes, such as maximum drawdown, rather than one historical path.

Key takeaways: Monte Carlo Simulation

• Monte Carlo resamples a backtest into thousands of paths to build an outcome distribution
• The percentile maximum drawdown, not the single historical one, should drive sizing
• Bootstrapping is more conservative than pure reshuffling and usually more honest
• It cannot invent tail events the sample never contained
• It measures variability, not whether the underlying edge is real

Monte Carlo Simulation at a glance

Monte Carlo Simulation in simple words

Your backtest gave you one equity curve, but that exact order of wins and losses was partly luck. Monte Carlo reshuffles or resamples those trades thousands of times to ask what else could plausibly have happened. The spread of results tells you how deep a drawdown you should be prepared for, not just the one your single backtest showed.

What Monte Carlo Simulation is for

Monte Carlo exists because a single backtest path understates uncertainty: it shows one realisation of a random process, and the worst drawdown you actually face is often deeper than the one lucky ordering in your historical run.

Monte Carlo Simulation — professional explanation

The two main resampling methods

The most common approach resamples at the trade level. You take the backtest's realised trade returns and either reshuffle their order (permutation, keeping the same set of trades but a new sequence) or draw them with replacement (bootstrap, allowing some trades to repeat and others to drop out). Each resample is compounded into a fresh equity path. A second approach resamples daily or bar returns rather than trades, which preserves position-level exposure but breaks the autocorrelation of a strategy that holds through trends. The choice of unit determines exactly which assumptions you are stress-testing.

What reordering versus bootstrapping tells you

Pure reordering (permutation) holds the multiset of outcomes fixed and varies only their sequence, so it isolates path risk: how much your drawdown depends on the order in which the same wins and losses arrive. Bootstrapping with replacement additionally varies which trades occur, so it also captures sampling risk: the chance that your historical trades were a favourable draw from a wider distribution. Bootstrapping generally produces a wider, more conservative outcome distribution, which is usually the more honest question for risk sizing.

Reading the output distribution

Run several thousand resamples and you get a distribution for each metric of interest. The headline is usually the distribution of maximum drawdown: the 5th and 95th percentiles bracket a plausible range, and the median maximum drawdown is often noticeably worse than the single historical drawdown, because your one backtest path was one relatively benign ordering. You can also read the probability of ruin (paths that breach a capital floor), the distribution of final equity, and the spread of Sharpe or CAGR. These percentiles, not the single backtest number, should drive position sizing and drawdown tolerance.

Key assumptions and where they break

Trade-level resampling assumes trades are exchangeable and roughly independent. That assumption fails when returns are serially correlated, when position sizing depends on recent equity (so a reshuffled sequence would never have been sized that way), or when trades cluster in regimes. Reshuffling a trend-following system that rides long winners can destroy the very autocorrelation that produced the edge, understating both its returns and its risk. Block bootstrapping, which resamples contiguous chunks of returns, partly preserves autocorrelation and is the standard fix, at the cost of fewer effective independent samples.

Monte Carlo on parameters and data, not just trades

Beyond resampling realised trades, Monte Carlo can perturb the inputs: jitter fill prices by a slippage distribution, randomly skip a fraction of trades to simulate missed signals, or add noise to the price series and re-run the whole strategy. This tests execution robustness rather than path luck: a strategy whose profit evaporates when fills are jittered by one tick, or when 5 percent of trades are randomly dropped, is dangerously fragile even if its trade-reshuffle distribution looks tolerable. Different Monte Carlo designs answer different robustness questions and should not be conflated.

What it can and cannot prove

Monte Carlo widens a single backtest into a distribution, which is its great strength for setting realistic drawdown expectations and position sizes. But it cannot create information that is not in the data: if your historical sample never contained a 2008- or 2020-scale shock, no reshuffling of it will invent one, so Monte Carlo systematically understates true tail risk when the sample lacks extremes. It also cannot detect look-ahead bias or overfitting in the underlying rules; it only characterises the variability of whatever the backtest already produced. It is a lens on uncertainty, not a source of new evidence.

Formula

How Monte Carlo Simulation looks visually

Worked example: Monte Carlo Simulation

Trade reshuffling vs Bootstrapping with replacement

Advantages of Monte Carlo Simulation

• Converts one lucky equity path into a full distribution of outcomes
• Gives a realistic, percentile-based maximum drawdown for position sizing
• Estimates probability of ruin against a capital floor
• Can also test execution robustness by jittering fills or dropping trades
• Requires no new data, only the backtest's existing trade or return series

Limitations of Monte Carlo Simulation

• Cannot invent tail events absent from the historical sample
• Naive reshuffling breaks the autocorrelation that trend strategies rely on
• Assumes trades are exchangeable, which fails under equity-dependent sizing
• Says nothing about overfitting or look-ahead bias in the underlying rules
• Percentiles are noisy if too few resamples are run

Why Monte Carlo Simulation matters in practice

• Reframes risk from a single number to a plausible range, improving sizing decisions
• Often reveals the historical drawdown was optimistic by a wide margin

How professionals treat Monte Carlo Simulation

Quant risk teams use Monte Carlo to set drawdown budgets and position sizes against a percentile of the outcome distribution rather than the single historical path, typically bootstrapping thousands of resamples and reporting the median and 95th-percentile maximum drawdown. Where autocorrelation matters they switch to block bootstrapping, and they complement trade-reshuffling with fill-jitter and trade-dropout runs to probe execution fragility. They are explicit that Monte Carlo widens the picture of known risk but cannot manufacture the tail events history did not sample.

Common misconceptions about Monte Carlo Simulation

Common mistakes with Monte Carlo Simulation

• Treating the single backtest drawdown as the worst case instead of a percentile of a distribution
• Reshuffling a serially correlated trend strategy and destroying its real behaviour
• Resampling trades that were sized from recent equity, an ordering that could never have occurred
• Assuming Monte Carlo captures crash risk when the sample contained no crash
• Running too few simulations, so the tail percentiles are unstable
• Believing Monte Carlo validates the strategy's edge when it only characterises variability

Monte Carlo Simulation: frequently asked questions

Voice search: how people ask about Monte Carlo Simulation

Natural-language questions people ask about Monte Carlo Simulation.

Sources & references

• Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). John Wiley & Sons.