How to read Monte Carlo simulation results for trading (2026)

A Monte Carlo simulation takes your strategy's trades and reshuffles them into thousands of alternative sequences, then reports the distribution of outcomes instead of a single number. Reading one comes down to four things: the percentile lines (P25/P50/P75/P90), the risk-of-ruin figure, the spread between the lines, and the difference between the median path and your one historical backtest. This guide walks through each, in the order you should check them.

Why isn't my backtest enough on its own?

Because your backtest is one draw from a distribution. The trades your strategy produced are real, but the order they arrived in is luck — and on a funded account with a drawdown limit, order is everything. The same 200 trades that pass an evaluation in one sequence can breach the same account in another, with identical totals. A Monte Carlo simulation makes that visible: it holds your trade population constant and varies only the sequence, which turns "my backtest made $12,000" into "here is the range of outcomes these trades can produce."

If the simulation resamples in blocks rather than single trades, it also preserves streakiness — losing runs stay clustered the way markets actually cluster them, instead of being shuffled into unrealistic smoothness.

What do P25, P50, P75 and P90 mean?

Percentiles rank the simulated outcomes from worst to best. P50 is the median: half the simulated runs ended below it, half above. P25 is the outcome a quarter of runs failed to reach — a "mediocre but plausible" scenario. P75 and P90 are progressively lucky runs.

How to use each line:

• P50 is your planning number. If the median outcome doesn't clear your goal — target reached, payout achieved, year survived — the strategy needs to change, not your hopes.
• P25 is your patience number. It tells you what a disappointing-but-normal stretch looks like, so you can recognize one without abandoning a working system.
• P90 is your marketing detector. When someone shows a single spectacular equity curve, you're usually looking at their P90, not their expectation.
• The number to plan around is never the best line. A strategy you'd only keep trading in its P75 world is a strategy you will abandon in the median world.

One trap: percentile lines are not paths any single account will follow. They are cross-sections of many paths. Your actual year will wander between them.

How do I read risk-of-ruin?

Risk-of-ruin (or breach probability) is the share of simulated runs that hit a defined failure line — on funded accounts, the trailing or fixed drawdown limit. Read it against the limit you actually trade under, not a generic account blow-up: a strategy with 2% risk of losing its full notional can still show 30% probability of clipping a $2,500 trailing floor, because the floor is far closer than zero.

Three practical rules. First, ruin probability is a function of position size before anything else — the same trades at half size can drop breach probability several-fold, which is why the useful direction of the calculation is backwards: choose an acceptable breach probability, then solve for size. Second, check what "ruin" was defined as in the tool you're reading; a simulation that measures distance-to-zero tells you little about a trailing floor. Third, remember that intraday-trailing accounts consume buffer through open-profit giveback, so a simulation built on closed trades understates breach risk on those rule sets.

What does the spread between percentiles tell me?

The gap between P25 and P75 is a direct reading of how much luck participates in your outcome at this sample size. A narrow band says the strategy's result is mostly determined by its edge; a wide band says sequence dominates. Two systems with identical medians and very different spreads are not equivalent — the wide one demands smaller size, more patience, and a longer evaluation window to prove anything.

The spread also shrinks with trade count. If the band looks unusably wide, the honest fixes are more trades (frequency, more instruments) or smaller size — not re-running the simulation until it looks better.

A worked reading, start to finish

Illustrative numbers. A simulation of 1,500 resampled sequences on a 50K account reports: P25 = +$2,100, P50 = +$6,400, P75 = +$11,200, breach probability 8%, median time-to-target 74 days. The reading: the median run clears a $3,000 evaluation target comfortably; a normal-bad year still finishes positive; roughly one account in twelve hits the floor before anything else happens, so running two accounts in parallel or cutting size is a legitimate discussion; and the plan should budget about three months to target, not the three weeks the luckiest paths show. That's the whole skill — every number answers a planning question, none of them predicts your specific year.

You can run this on your own trade list, with block resampling and a trailing-floor breach definition, in the Monte Carlo simulator; if the question is specifically "will these trades pass this firm's rule set", the pass estimator frames the same distribution against a chosen program's rules. Before trusting any of it, make sure the input backtest itself holds up — the backtest quality checker covers the usual ways a trade list lies.

FAQ

Is P50 the result I should expect?

It's the single best planning number, but "expect" is too strong: half of all runs finish below it. Plan around P50, size around P25 and the breach probability, and treat anything above P75 as luck you don't get to schedule.

What is a good risk-of-ruin number?

There's no universal threshold — it's a budget decision. On evaluation accounts many systematic traders work backwards from single-digit breach probability per attempt and let position size be the output. A breach probability above ~20% means the account fee is closer to a lottery ticket than an investment.

Why do my Monte Carlo results differ between tools?

Different resampling (single-trade vs block), different run counts, different ruin definitions, and whether costs and the trailing mechanics are modeled. Compare settings before comparing outputs — the disagreement is usually in the assumptions, not the math.

Does Monte Carlo predict future performance?

No. It maps the range of outcomes your historical trades can produce under resequencing. If the strategy's edge decays, no simulation of old trades will warn you — that's what forward tracking is for.

---

Illustrative figures throughout; simulation outputs are modeled results, not live performance. Past performance does not guarantee future results. Not financial advice.