Monte Carlo Simulation in Trading: Stress-Testing Your Strategy

How Monte Carlo Works

Step 1: Take all trades from your backtest (e.g., 247 trades with their individual P&L values).

Step 2: Randomly shuffle the order of these trades.

Step 3: Calculate a new equity curve from this shuffled sequence.

Step 4: Record the max drawdown, final balance, and other metrics from this run.

Step 5: Repeat 1,000–10,000 times.

The result: a distribution of possible outcomes, all using your ACTUAL trade results but in different orderings.

How Many Simulations Do You Need?

The number of iterations directly affects the reliability of Monte Carlo results. Too few runs produce unstable distributions that change every time you re-run the analysis. Too many waste computation time without meaningfully improving accuracy.

500 iterations: Rough estimate only. Percentile values can shift by 2–5% between runs. Adequate for quick sanity checks.

1,000 iterations: The practical minimum. Distribution shapes stabilize, percentile values are repeatable within 1–2%.

5,000–10,000 iterations: Professional-grade. Tail percentiles (1st, 99th) are well-populated, giving reliable worst-case estimates.

What Monte Carlo Reveals

The backtest showed 22% max drawdown. But Monte Carlo reveals that the same trades in a worse ordering could produce 38% drawdown. If you can only handle 25% drawdown, your strategy has a significant probability of exceeding your tolerance — even though the historical backtest looked comfortable.

Interpreting Confidence Intervals

Monte Carlo's most valuable output is the confidence interval for each metric. Understanding how to read these intervals separates casual users from those who extract real insight:

The 50th percentile (median) represents the most likely outcome — usually close to the single-backtest result but not identical.

The 5th percentile is your planning number — the worst realistic outcome. Use this for capital reserves, position sizing, and stop-trading thresholds.

The 95th percentile is the optimistic boundary. If your plan depends on achieving this level, you're building on hope rather than probability.

The spread between 5th and 95th percentiles measures strategy stability. A spread of 15–20% is predictable; 10–55% is a lottery ticket with wildly different outcomes depending on trade ordering.

Monte Carlo for Drawdown Estimation

Drawdown estimation is where Monte Carlo provides the most actionable insight. A single-path backtest gives you the historical maximum drawdown, but this number is a single observation from one specific sequence. Monte Carlo generates a full drawdown distribution:

From the example above, the Monte Carlo distribution might show: 10% of simulations exceeded 30% drawdown; 5% exceeded 35%; 1% exceeded 42%. This drives two decisions:

Position sizing: If your maximum tolerable loss is $20,000 and the 5th percentile drawdown is 35%, your maximum allocation is $20,000 / 0.35 = $57,143 — versus $90,909 if you sized from the historical 22%.

Stop-trading rules: If drawdown exceeds the 1st percentile, something has fundamentally changed. Setting a threshold at this level protects capital from regime changes the backtest didn't capture.

Why Trade Order Matters

Consider a simple example: 10 trades with 5 wins of +$200 and 5 losses of −$200. The final result is always $0. But the equity curve varies dramatically:

Best ordering: W, W, W, W, W, L, L, L, L, L → Peak at +$1,000, drawdown at end = $1,000

Worst ordering: L, L, L, L, L, W, W, W, W, W → Trough at −$1,000, recovery at end

Average ordering: Mixed → moderate drawdown throughout

The final P&L is identical. The EXPERIENCE is completely different. Monte Carlo shows the full range of possible experiences.

Portfolio-Level Monte Carlo

Monte Carlo analysis becomes even more powerful when applied to multiple strategies simultaneously. A portfolio of three uncorrelated strategies has trade sequences that interact in complex ways. One strategy's drawdown period might overlap with another's winning streak — or with another's drawdown, compounding losses.

Portfolio-level simulation shuffles each strategy's trades independently, then combines them into a portfolio equity curve. After thousands of iterations, you can answer questions single-strategy analysis cannot: What is the probability that all strategies draw down simultaneously? Does adding a fourth strategy meaningfully reduce worst-case portfolio drawdown, or is diversification already saturated?

The combined drawdown profile is not the sum of individual strategy drawdowns — it depends on timing interactions that only simulation can reveal.

Practical Applications

Setting realistic drawdown expectations: Use the 5th percentile max drawdown from Monte Carlo, not the single backtest drawdown, to set your stop-trading threshold.

Position sizing: Size your positions so the 5th percentile max drawdown in dollar terms is survivable. If Monte Carlo shows 35% max drawdown at the 5th percentile, and you can handle $17,500 max loss, your maximum account allocation is $50,000.

Strategy comparison: When comparing two strategies, Monte Carlo distributions are more meaningful than single backtest numbers. A strategy with lower average return but tighter confidence intervals may be superior to one with higher average return but wide, unpredictable ranges.

Limitations

Monte Carlo assumes trade independence — each trade's result is independent of previous trades. This isn't always true in real markets: trending markets produce clustered wins, and choppy markets produce clustered losses. The simulation may underestimate serial correlation in trade outcomes. Despite this limitation, Monte Carlo provides vastly better risk estimates than a single backtest path.

StratBase.ai provides comprehensive backtest metrics that help you understand your strategy's risk profile — setting realistic expectations before live trading.

Further Reading

• RSI on Investopedia
• Drawdown on Investopedia