Monte Carlo: a thousand futures, not a forecast.
Imagine running the same trading idea thousands of times. Same starting money, win rate and payoff—but different sequences of wins and losses. The results explain why a good average does not guarantee a smooth journey.
Why do traders use it?
Suppose your strategy wins 45% of trades, earning 1.5 times what you risk on winners. You might calculate a positive expected outcome and conclude that your capital should grow. But how rough could the journey be before it grows? Could a sequence of losses force you to stop trading?
Monte Carlo tests the consequences of those assumptions across many hypothetical paths. It does not discover the true win rate, and it cannot predict the next trade. It is a stress test of a model you chose.
How to read the results.
P10, median and P90
At the final trade, P10 is the balance below which about 10% of simulations end. P90 has about 10% above it. The median splits the sample in half. Neither end is a worst or best case.
The fan chart shows cross-sectional percentiles by trade, not three individual portfolios.
Maximum drawdown
For each path, we track the largest decline from its own earlier equity peak. We then report the median and 90th-percentile maximum drawdown across paths.
A 30% drawdown from a peak is not the same as falling to 70% of starting capital.
Chance of crossing a floor
We count a path once if its equity ever touches your chosen fraction of starting money. Recovering later does not undo the event. The simulator also shows an exact benchmark for this special two-outcome model.
A 95% Wilson interval only measures randomness from sampling paths—not uncertainty in your strategy assumptions.
Two trades. One surprising lesson.
You start with $10,000. You risk 2% of whatever remains in the account on each trade, and winning earns 1.5R. That is +3% for a win or −2% for a loss.
Notice the subtlety: under this fixed-fraction two-outcome model, final capital depends on the total number of wins and losses, not on their order, because multiplication commutes. But maximum drawdown and whether a capital floor was crossed on the way absolutely can depend on the order. That is why tracking just average or final returns can hide important risk.
Real trades rarely follow perfect +3% and −2% outcomes. Spread, commissions, slippage, gaps, changing R multiples, clusters of losses and market regimes can make the observed distribution significantly different.
Random sample versus exact probability.
Monte Carlo works by sampling many paths with a reproducible pseudo-random seed. A few thousand paths produce a useful illustration, but the percentage of paths that cross a floor changes slightly when you change the seed. More paths reduce sampling variation—they do not repair wrong probabilities, incomplete payoff distributions or missing execution costs.
For the specific binary fixed-fraction assumptions used in both Risk Lab tools, the chance of first crossing a starting-capital floor also has an exact finite-horizon dynamic-programming calculation. We show this number beside the sampled estimate. The difference is sampling noise, not disagreement about market conditions. The underlying assumptions are equally strong in both.
The algorithm uses the same win and loss multipliers for each trade. It does not infer them from a Bitcoin chart. Nor does a positive mathematical expected equity tell you that the median simulated account will grow.
What Monte Carlo cannot establish.
A strategy that performed well on 50 trades does not automatically have a known win probability. Small samples, overfitting and shifts in volatility make those estimates uncertain. Our current binary model also excludes correlations between trades, partial exits, fees, funding, liquidation, margin constraints and losses beyond a planned stop.
For more realistic simulations, a later research engine can take distributions of actual trade outcomes supplied by the user. But even resampling historical trades would not guarantee that the future follows the same distribution.
Therefore, use this tool to ask “what happens under these assumptions?”—not “what will happen to my account?”
Change one assumption at a time.
Compare 1% and 5% risk per trade with the same win rate and R. See how the median, downside percentiles and drawdowns change.
Method and further reading
Our reproducible calculation, formulas, inputs and limitations are explained in the methodology section and the companion Risk of Ruin guide. This article uses original educational examples rather than third-party exchange data.