Monte Carlo Thinking
From portfolio-level intuition to company-level distributions: why a single-number valuation hides the shape of the payoff, and how distribution thinking drives position sizing.
Why simulate?
Financial markets are complex systems where analytical solutions are often impossible. The number of interacting variables, feedback loops, and non-linear relationships makes closed-form calculation impractical. Monte Carlo simulation offers a different approach: instead of solving for the answer, you generate thousands of possible futures and observe the distribution of outcomes.
This is about understanding the range of what could happen — and how likely different regions of that range are, admitting the truth that forecasting is hard.
Building intuition for uncertainty
The fundamental insight of Monte Carlo is that a single point forecast is almost always wrong. What matters is the distribution: the central tendency, the spread, the skew, and the tails. A portfolio with a 7% expected return and 18% volatility can produce a wide range of outcomes over 20 years — from wealth destruction to substantial wealth creation.
Use the simulator below to explore how mean return, volatility, and time horizon interact. Each click of “Simulate” generates a fresh set of 200 possible paths — notice how the fan of outcomes widens over time, and how the median and mean diverge.
Monte Carlo Simulation
Simulate 200 wealth paths using geometric Brownian motion. Adjust the parameters and click Simulate to see how return assumptions and volatility shape the range of outcomes.
Even with positive expected returns, the range of outcomes is wide. Higher volatility dramatically increases the spread between best and worst paths. This is why time horizon matters so much — longer horizons allow the law of large numbers to work in the investor's favour, but never eliminate the possibility of poor outcomes entirely.
From portfolios to individual companies
The portfolio-level simulation above is the intuition pump. It shows that even a simple wealth path hides a wide distribution of outcomes. But the deeper application of Monte Carlo thinking is at the level of an individual business.
A traditional valuation — whether built bottom-up from cash flows or anchored on multiples of earnings — produces a single answer: fair value. That answer is the product of dozens of assumptions about a company’s future: growth rates, profitability, reinvestment needs, the multiple it deserves, the return you require. Each of those inputs is itself uncertain. Propagate that uncertainty through the model and you do not get a number. You get a distribution.
The shape of that distribution is the actual object of analysis. Two companies can have the same point-estimate fair value and completely different payoff profiles. A regulated utility with a narrow cone of outcomes is a fundamentally different investment from a growth tech name with a long right tail — even if both models pop out the same £20 per share.
Why does this matter? Because the distribution of returns is what you actually live through as an investor — not the expectation. This is the core insight from ergodicity: ensemble averages can be misleading when you are sampling a single path through time. A 50/50 coin flip with a positive expected value can still wipe you out, because the order of outcomes and the shape of the distribution determine your lived experience of the payoff, not its mean.
Four companies, four distributions
To see how different the shape of the payoff can be across different kinds of business, try this simulator. It runs 3,000 Monte Carlo simulations, with log-normal growth and a re-rated exit multiple in each path, and plots the resulting outcome distribution against the current price. The five archetypes use different growth, margin, volatility, and re-rating assumptions; the engine is identical.
Five companies, five distributions
The same Monte Carlo engine, five different company archetypes. Each simulation draws log-normal growth paths, margin variance, and a re-rated exit multiple — the three ingredients that make realised equity outcomes right-skewed. Pick an archetype to see how the shape of the outcome distribution changes — and how much more that shape tells you than any single-point estimate.
Revenue path — 5-year cone
5-year outcome distribution — £ per share
An extremely narrow cone and a tight outcome distribution. Most paths cluster in a small range around the median. Regulated economics bound both the downside and the upside. A single-point forecast almost works here — but a low-variance outcome is also a low-return one.
Notice how the shape of the distribution, not the median, distinguishes the five businesses. The utility is almost a point estimate. The compounder builds a tight distribution through reliable growth. The growth tech case is a lottery ticket with genuine downside. The cyclical has a wide, symmetric spread. The mature value case tilts right of the current price, but retains real left-tail risk if execution deteriorates.
Kelly, position sizing, and the shape of the bet
If the output of your analysis is a distribution, the natural next question is: how much do I bet? The Kelly Criterion gives a principled answer. Kelly tells you to size positions in proportion to your edge relative to the variance of the payoff. Higher expected return, larger position. Higher variance, smaller position. Kelly explicitly penalises distributions with fat left tails — exactly the behaviour you want.
The practical problem is that classical Kelly assumes you know the distribution. In reality you are estimating it. A Monte Carlo simulation of a company is a structured way of producing that estimate: explicit assumptions, explicit variance, explicit percentiles. It lets you size positions using the full shape of the payoff rather than a point forecast and a gut feel about uncertainty.
The point is not to apply the formula mechanically. Sizing decisions should scale with the shape of the distribution you believe in, not with your conviction in a single number.
Practical applications
Monte Carlo thinking is useful wherever company-level uncertainty is irreducible:
- •Valuation: replacing a point estimate with a distribution of fair values — and seeing which assumptions drive most of the variance.
- •Position sizing: feeding the shape of the payoff into a Kelly-style framework, sizing down when the left tail is fat.
- •Scenario planning: stressing individual drivers (revenue growth, margin, terminal multiple) and seeing how the distribution of equity value shifts.
- •Conviction calibration: distinguishing a 'strong buy' where the distribution is tight and offset from a 'strong buy' where the median is high but the left tail is real.
- •Communicating uncertainty: fair-value fan charts and histograms are a powerful way to show stakeholders that the future of a business is a distribution, not a target price.
The discipline is not in the simulation itself — it is in forcing yourself to think in distributions rather than forecasts, and to make decisions that are robust across the range of outcomes rather than optimal for a single predicted path.
Key Takeaway