Monte Carlo simulation estimates outcomes by running many random trials instead of solving equations. You define a model (inputs, relationships, and distributions), draw random values for uncertain inputs repeatedly, and observe the distribution of results. The method was developed at Los Alamos during the Manhattan Project; Stanisław Ulam and John von Neumann named it after the casino. When a system is too complex for closed-form solution or when you need the full distribution of outcomes — not just a point estimate — Monte Carlo delivers.
The core move: replace "what is the answer?" with "what do 10,000 runs of this process look like?" You get a histogram of outcomes, percentiles (e.g. 10th, 50th, 90th), and the chance of crossing a threshold (e.g. probability of ruin, or of hitting a target). That supports deciding under uncertainty: you see the range of plausible futures, not a single number that hides variance. In business, common uses include project completion times, revenue under multiple drivers, portfolio returns, and the impact of correlated risks.
The discipline is specifying the model honestly. Garbage in, garbage out applies. If you assume normal distributions when tails are fat, or ignore correlation between inputs, the simulation will understate risk. The strength of Monte Carlo is that it forces you to make assumptions explicit — each input has a distribution — and then it shows you what those assumptions imply when combined. Use it when decisions depend on the shape of the outcome distribution, not just the mean.
Section 2
How to See It
Monte Carlo reveals itself when someone runs thousands of random trials to quantify risk or range of outcomes instead of using a single scenario or a simple formula. Look for tools like @RISK, Crystal Ball, or custom scripts; for discussions of "10,000 runs," "percentile outcomes," or "probability of exceeding X." The diagnostic: are we characterising the full distribution of an outcome by sampling from uncertain inputs?
Business
You're seeing Monte Carlo Simulation when a CFO models revenue by drawing random growth rates, conversion rates, and churn from specified distributions and running 10,000 years of the business to get a distribution of outcomes. The output might be "30% chance we hit $50M, 50% we're between $40M and $60M, 10% we're below $30M." Decisions then use the full distribution, not a single forecast.
Investing
You're seeing Monte Carlo Simulation when a portfolio manager simulates asset returns and correlations over many paths to estimate the distribution of portfolio value in one year or to compute value-at-risk. The simulation captures path-dependence and correlation that a simple expected-return calculation would miss.
Projects
You're seeing Monte Carlo Simulation when a project lead estimates completion by simulating each task's duration (with min/mode/max or a distribution) and summing along the critical path across many runs. The result is a distribution of project end dates and the probability of finishing by a given date.
Risk
You're seeing Monte Carlo Simulation when an insurer or risk team simulates claim frequency and severity with random draws to estimate the distribution of total losses. Capital and pricing decisions use the tail of that distribution, not just the mean.
Section 3
How to Use It
Decision filter
"Before committing to a plan that depends on uncertain inputs, ask: do I need the full distribution of outcomes or just a point estimate? If the former, build a simple model, assign distributions to key inputs, run thousands of trials, and decide using percentiles and tail probabilities — not the average path."
As a founder
Use Monte Carlo for runway, revenue, and launch timelines. Model key drivers (e.g. burn, growth, conversion) with ranges or distributions; run many trials. Look at the 25th and 75th percentiles, not just the "base case." That reveals how sensitive outcomes are to variance and correlation. Avoid the trap of a single spreadsheet scenario that hides tail risk. When raising or planning, show the distribution: "We have an 80% chance of reaching 18 months of runway under these assumptions."
As an investor
Simulate portfolio or strategy outcomes when payoff depends on the shape of the distribution — e.g. probability of a 50% drawdown or of missing a liquidity need. Use it to stress-test assumptions: what correlation or fat tail would break the thesis? The value is in seeing the full range, not in trusting a single NPV or IRR.
As a decision-maker
When a team presents a single forecast, ask for the distribution. If they haven't run simulations, request a simple Monte Carlo: identify the 3–5 uncertain inputs, assign plausible ranges or distributions, run 1,000+ trials. Decisions that depend on "what if we're wrong?" should use percentiles and tail probabilities, not just the mean.
Common misapplication: Treating the simulation output as truth. The output is conditional on the model and the input distributions. If those are wrong, the distribution is wrong. Use sensitivity analysis: vary key assumptions and see how the outcome distribution shifts.
Second misapplication: Using Monte Carlo when a closed-form or single-scenario answer is enough. If the decision is robust to the full range of inputs, simulation may be unnecessary. Reserve Monte Carlo for when the distribution shape or tail risk drives the decision.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
Ed ThorpMathematician, Beat the Dealer; hedge fund founder
Thorp used probabilistic methods and simulation-like reasoning to price warrants and manage risk. His approach was to characterise the distribution of outcomes and then size positions and hedges accordingly — the same spirit as Monte Carlo: know the full distribution, not just the expected value.
Renaissance uses massive data and statistical modelling; simulation and distributional thinking are central. Simons has emphasised understanding the full distribution of returns and the risk of rare events, not just average returns — the mindset that Monte Carlo formalises.
Section 6
Visual Explanation
Monte Carlo Simulation — Draw random inputs from their distributions, run the model many times, and inspect the distribution of outputs. Percentiles and tail probabilities inform the decision.
Section 7
Connected Models
Monte Carlo simulation sits at the intersection of probability, scenario thinking, and decision-making. The models below either reinforce its logic, extend it, or create tension with how we use it.
Reinforces
Probability Theory
Probability theory provides the distributions and the law of large numbers that justify Monte Carlo: the empirical distribution converges to the true distribution as trials grow. Simulation is applied probability — you encode beliefs as distributions and derive implications.
Reinforces
Scenario Analysis
Scenario analysis explores a few hand-picked futures (e.g. base, bull, bear). Monte Carlo generalises that by sampling many futures from specified distributions. Both force explicit assumptions; Monte Carlo adds the full distribution instead of a small set of scenarios.
Tension
Ludic Fallacy
The ludic fallacy is mistaking the clean randomness of games for the messy uncertainty of life. Monte Carlo can encourage that if you feed it tidy distributions that understate tail risk or correlation. The tension: use simulation to see the distribution, but don't assume your model captures reality.
Tension
Fermi Problem
Fermi problems use rough estimates and orders of magnitude for quick bounds. Monte Carlo is more precise but requires full distributions. The tension: when you lack data, Fermi-style bounds may be more honest than a Monte Carlo built on guessed distributions.
Section 8
One Key Quote
"Using the Monte Carlo method, one could trace the fate of a single neutron, or even of a single collision, and then average over many such trajectories."
— Stanisław Ulam, Adventures of a Mathematician
The idea: follow one random path, then repeat. The average (or the full distribution) over many paths is what you use. Ulam's formulation applies to any process you can simulate — one run is one draw from the outcome distribution; many runs reveal the distribution. The discipline is running enough trials and interpreting the result as a distribution, not a single trajectory.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Monte Carlo is underused in strategy and planning. Most teams rely on a single forecast or a handful of scenarios. That hides variance and tail risk. A simple simulation — 5–10 uncertain inputs, 1,000–10,000 runs — surfaces the range of outcomes and the probability of extremes. The cost is low; the benefit is seeing what you're actually betting on.
The value is in the percentiles, not the mean. The average of 10,000 runs is often close to a deterministic "base case." The 10th and 90th percentiles are where decisions change. Can we survive the 10th percentile? Do we need to plan for the 90th? Run the simulation and answer with numbers.
Specify distributions honestly. If you use narrow ranges or normal distributions when reality is fat-tailed, the simulation will understate risk. When in doubt, widen the ranges and run again. The goal is to capture plausible uncertainty, not to make the output look good.
Use it to stress-test narratives. When someone says "we'll be fine," ask: what distribution of outcomes does that assume? Run a Monte Carlo with their assumptions and show the tail. If the 25th percentile is ruin, the narrative is fragile.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A team models next year's revenue by assuming 10% growth and a single churn rate.
Scenario 2
A CFO runs 5,000 trials of cash flow, drawing burn rate and revenue from specified ranges, and reports '70% chance we have 12+ months runway.'
Scenario 3
A project manager adds 20% buffer to each task and sums to get the end date.
Scenario 4
An investor simulates 10,000 paths of portfolio returns using historical means and covariances, then checks the 5th percentile outcome.
Section 11
Summary & Further Reading
Summary: Monte Carlo simulation estimates the distribution of outcomes by running many random trials from specified input distributions. Use it when the decision depends on the full range or tail of outcomes, not just a point estimate. Build a simple model, assign distributions to key uncertainties, run thousands of trials, and decide using percentiles and tail probabilities. The output is only as good as the model and the input distributions — specify them honestly and stress-test with sensitivity analysis.
Shows how to assign distributions to uncertain quantities and use Monte Carlo to quantify and reduce uncertainty. Applied to business and project decisions.
Rigorous treatment of Monte Carlo and variance reduction in finance. For those building or auditing simulation models.
Leads-to
Sensitivity Analysis
After running Monte Carlo, sensitivity analysis identifies which inputs drive the output most. Vary one input at a time (or use correlation of input with output across runs) to see what matters. Simulation gives the distribution; sensitivity tells you where to refine the model.
Leads-to
Expected Utility Theory
Expected utility uses the full distribution of outcomes, weighted by probability. Monte Carlo supplies that distribution. The link: once you have the outcome distribution from simulation, you can apply expected utility (or other decision rules) to choose under uncertainty.
