PromptDesk Pro
← Back to the blog
How to validate your trading strategy with Monte Carlo simulation
You have 80 trades logged, the equity curve points up and the win rate reads 58%. Strategy validated? No. All you know is that one particular sequence of those trades ended well. The important question is a different one: what would have happened if those same trades had arrived in a different order?
That's exactly what a Monte Carlo simulation answers, and it's the difference between "it went well" and "I have an edge".
What Monte Carlo means in trading
The idea is simple: take your real trades (the individual P&Ls) and resample them at random thousands of times, building thousands of alternative equity curves with your same results in different orders. From that distribution come answers a single history can't give you:
- Distribution of final results: in the worst 5% of scenarios, do you end up winning or losing?
- Probable drawdown: your historical drawdown was $2,000 — but what's the 95th percentile drawdown when the losing streak clusters?
- Probability of finishing positive if you repeat this trading.
- Risk of ruin: probability of touching a loss threshold you can't come back from (or that violates your prop firm).
Why win rate misleads
Two classic problems Monte Carlo exposes:
1. The order of your trades is luck
A system with a 55% hit rate can chain 8 straight losses without ceasing to be the same system (over 100 trades, it's more likely than you think). If that streak arrives at the start — when your prop firm account has the trailing drawdown glued to it — it eliminates you. If it arrives at the end, you barely notice. Same system, same trades, opposite fate. Your real history only showed you one of those worlds.
2. Few trades = almost any result is noise
With 20 or 30 trades, the difference between "real edge" and "lucky streak" is statistically almost indistinguishable. Monte Carlo doesn't fix that by magic, but it shows it to you without mercy: if with your 25 trades the simulation says 35% of scenarios end in a loss, your "winning strategy" hasn't proven anything yet.
An example with numbers
Say you have 60 real trades: 55% win rate, average win $220, average loss $180, net result +$2,400, historical drawdown $1,100. Looks solid. We run 1,000 Monte Carlo simulations:
| Metric | Your history | Monte Carlo (1,000 scenarios) |
|---|---|---|
| Final result (median) | +$2,400 | +$2,350 |
| Worst 5% of scenarios | — | −$310 |
| Median drawdown | $1,100 | $1,350 |
| 95th percentile drawdown | — | $2,600 |
| Probability of finishing positive | — | 91% |
The honest reading: the edge looks real (91% of scenarios positive), but the drawdown "to expect in a bad scenario" is more than double the one you saw. If you were about to run this system on an account with a $2,500 trailing drawdown... you just discovered that 1 in 20 times it burns your account without the system having stopped working.
The practical rule: size your risk to survive the Monte Carlo 95th percentile drawdown, not the drawdown you saw in your history. The one you saw was luck; the 95th percentile one is the one you have to survive.
How to run a Monte Carlo simulation, step by step
- Gather your per-trade P&Ls (a minimum of 30–50 trades for it to mean anything; more is better).
- Resample with replacement: build a new sequence by picking trades at random from your list (the same trade can appear multiple times).
- Build the equity curve for that sequence and record: final result and maximum drawdown.
- Repeat 1,000+ times.
- Analyze the distribution: final-result percentiles, drawdown percentiles, % of positive scenarios, % that touch your ruin threshold.
You can do it in a spreadsheet with patience, or with a script. The limitation to know about: resampling assumes your trades are independent of each other — if your trading has strong sequential dependence (for example, you size up after losing), the simple model underestimates it.
Monte Carlo with your CSV, no formulas
The PromptDesk Pro Monte Carlo validator runs thousands of simulations on your real NinjaTrader history: result distribution, drawdowns by percentile, probability of profit and risk of ruin against the threshold you define (for example, your prop firm's trailing drawdown). Everything is calculated in your browser — your trades never leave your device.
Validate my strategy →Which decisions it changes (concrete cases)
- Before paying for a prop firm evaluation: if the 95th percentile drawdown exceeds the allowed trailing, don't pay yet — cut size or improve the strategy.
- To choose position size: combine Monte Carlo with the Kelly criterion; the Kelly fraction gives you the theoretical ceiling and Monte Carlo shows you what that ceiling feels like in drawdowns.
- To detect overfitting: if your backtest looks perfect but Monte Carlo shows high dispersion, the result depended on the order — a bad sign.
Related: Risk of ruin: the metric that decides whether you're still trading next year and how prop firm trailing drawdown works.
This content is educational and does not constitute financial advice. No simulation guarantees future results; Monte Carlo estimates scenarios from past data, which may not repeat. Trading involves risk of loss.