Monte Carlo simulation in portfolio analysis: modelling uncertainty across thousands of outcomes
Monte Carlo simulation applies repeated random sampling to model the range of possible outcomes for a portfolio. Where a single backtest produces one historical return path, a Monte Carlo simulation generates thousands of alternative paths—each drawn from the same statistical properties as the historical data—and analyses the full distribution of results. The technique does not predict what will happen; it quantifies the uncertainty around what might happen, making it one of the most honest tools available for long-run portfolio planning.
What Monte Carlo simulation does
In a standard portfolio Monte Carlo, the simulation takes the portfolio's estimated return distribution—its mean, volatility, and often its correlations with other assets—and uses random number generation to draw return sequences that are statistically consistent with those properties. Each simulation run produces a complete portfolio path from start date to end date. Running 10,000 such paths produces a distribution of end values, drawdowns, and recovery times. The analysis then reports percentile outcomes: the 5th percentile outcome, the median, the 95th percentile, and so on.
This approach is particularly useful for questions that historical backtesting cannot answer: what does the distribution of 20-year portfolio outcomes look like? What is the probability of a 40% drawdown occurring at any point in a 10-year investment horizon? What is the likelihood of the portfolio recovering within two years of a severe drawdown? A single historical path answers these questions with one data point; Monte Carlo provides thousands.
Monte Carlo versus backtesting
Backtesting uses actual historical returns in their historical sequence. It shows what the portfolio would have done over a specific past period. Monte Carlo simulation draws from the statistical properties of returns but generates new sequences—so it can produce scenarios that never occurred in history, including sequences of bad returns that cluster in ways the historical record has not yet produced. This is both its strength and its limitation: it explores scenarios beyond the historical sample, but it can only be as accurate as the statistical model used to generate the random returns.
The two approaches are complementary. Backtesting tests the strategy against what actually happened; Monte Carlo tests the portfolio's risk characteristics against a broader range of plausible futures. Both are subject to the limitation that the future may not resemble the past—but Monte Carlo at least quantifies the range of outcomes under the assumed model, rather than treating one historical episode as the only possible reference. See backtesting investment strategies and overfitting in quantitative investing for further context.
Assumptions and their consequences
Monte Carlo results are only as reliable as the statistical model used to generate the returns. If the simulation assumes normally distributed returns, it will underestimate the frequency of extreme outcomes for assets with fat-tailed return distributions—such as equities, commodities, and cryptocurrencies. A more sophisticated simulation will draw returns from a distribution that matches the observed skewness and kurtosis of the historical series, or will use a block bootstrap method—resampling chunks of actual historical returns—to preserve the autocorrelation structure of real market behaviour.
Correlation assumptions matter too. If the simulation models assets as having constant pairwise correlations, it will not capture the well-documented tendency for correlations between asset classes to rise sharply during market crises—precisely when diversification is most needed. A Monte Carlo that ignores regime changes in correlation will systematically underestimate tail risk in multi-asset portfolios.
Limitations
Monte Carlo simulation is a model of a model. The outputs reflect the assumptions built into the simulation, not the actual future. Running 100,000 simulations does not reduce the uncertainty about future returns—it only maps the uncertainty implied by the model's assumptions more precisely. Users who treat Monte Carlo output as predictions rather than scenario distributions will draw false confidence from the results.
The technique also struggles with structural changes in the investment environment. If the simulation is calibrated on a period of declining interest rates, it will not produce scenarios consistent with a sustained period of high rates—yet that scenario is relevant for any forward-looking analysis. Monte Carlo should be used alongside stress testing and scenario analysis, not as a substitute for them.
Monte Carlo simulation in pfolio
Monte Carlo simulation is not currently a built-in feature in pfolio. The platform's risk framework uses historical time-series statistics—volatility, drawdown, expected shortfall, value at risk—rather than simulation-based projections. Investors who want Monte Carlo analysis can use pfolio's reported statistics (mean, volatility, correlation) as inputs to external simulation tools.
Related articles
Disclaimer
Get started now
