Monte Carlo Simulation: Stripping Luck Out of Your Backtest

The cumulative return curve from a backtest is just one path drawn in the order trades happened. With the same win rate and average P&L, a different trade order produces a different max drawdown and CAGR. Monte Carlo simulation strips out that path-specific luck.

Two Main Approaches

Approach	What it does	Goal
Trade shuffle	Reorder observed trade P&Ls	Show the spread of possible max DDs
Parametric sampling	Sample from the P&L distribution	Estimate worst-case drawdown magnitudes

Shuffling preserves win rate and trade count, so interpretation is simple. Parametric sampling assumes the future matches the past’s distribution, but gives a clearer picture of “plausible worst cases.”

What to Measure

Run 1,000 paths and you get a distribution of drawdowns and returns. Useful readings:

• 95th percentile max DD: drawdown that’s exceeded 1 in 20 times
• 5th percentile final return: the “bad luck” lower bound
• Bust probability: fraction of paths that drop below your capital floor

Making decisions from a single backtest result is prone to confusing a lucky path for skill.

Practical Notes

• Trade independence is an approximation. Real trades correlate (same regime pushes multiple names). Shuffling ignores this, often producing optimistic DD estimates
• Sample size matters. Fewer than ~50 trades gives thin diversity in shuffles
• Regime dependence. A distribution estimated from bull-only data will mislead about the future

Relationship to Walk-Forward

Monte Carlo captures path luck; walk-forward captures period luck. Using both gives a more three-dimensional view of robustness. See our walk-forward primer.

Try It in QuanTest

Run a backtest in QuanTest, then use the trade log to run Monte Carlo for DD distribution. Start with 1,000 shuffles of trade order and observe the spread.

---

This article is for educational purposes. It does not guarantee the profitability of any strategy or future performance. Investment decisions are your own responsibility.