Monte Carlo Analysis

What This Page Is For

Use this page to understand what the Monte Carlo robustness tests measure.

Trade-Order Shuffling

This method uses the existing backtest trades and randomly changes their order.

Use it to ask:

• did the strategy depend on a lucky sequence of trades?
• would drawdown look worse if wins and losses arrived in a different order?

It is fast because it does not rerun the strategy on candles.

Candles-Based

This method creates synthetic OHLCV market paths from the original candle history and reruns the full backtest for each path.

For each scenario, BEC:

1. calculates historical close-to-close returns;
2. samples those returns in a new order, with repetition;
3. adds small statistical noise equal to 15% of historical return volatility;
4. rebuilds valid Open, High, Low, Close, and Volume candles;
5. runs the selected strategy again on that synthetic path.

It does not edit the original candles. It creates new DataFrames for each scenario.

Robustness Score

The MC Score is a 0-100 robustness score. It compares the original result with the Monte Carlo scenario distribution.

The score uses three components:

Component	Weight	What It Measures
Median return	55%	Whether the typical Monte Carlo scenario keeps returns close to the original backtest
Worst 5% return	25%	Whether the lower-tail scenarios still hold up
Drawdown	20%	Whether the worst 5% drawdowns are controlled versus the original drawdown

The final score is adjusted by the valid scenario ratio:

MC Score =
(
  median return component * 55%
  + worst 5% return component * 25%
  + drawdown component * 20%
)
* valid scenarios / total scenarios

For example, if BEC runs 200 scenarios and all 200 are valid, the valid scenario ratio is 1.0. If only 120 of 200 are valid, the score is multiplied by 0.6.

Components

The median return component compares the median scenario return with the original return. If the median scenario return is much lower than the original backtest, the MC Score drops.

The worst 5% return component looks at the lower tail of the scenario distribution. This penalizes strategies that only look strong in the original path but break down in weaker synthetic paths.

The drawdown component compares the original drawdown with the worst 5% scenario drawdown. If Monte Carlo scenarios create much larger drawdowns than the original backtest, the score drops.

All components are capped between 0 and 100 before weighting.

Common interpretations:

Interpretation	Rule	Meaning
Robust	Score >= 75	Scenario results remain close enough to the original
Moderate robustness	Score >= 55 and < 75	Usable signal, but not strongly stable
Sequence-sensitive	Score < 55 for trade-order shuffling	Trade order has a large effect
Market-path fragile	Score < 45 for candles-based tests	Synthetic market paths degrade the result heavily
Insufficient scenarios	Too few valid scenarios or insufficient trade sample	The result should not be trusted

The MC Score is not the same as the Backtesting Strategy Quality Score. MC Score is only about robustness under Monte Carlo scenarios; Strategy Quality Score combines return, risk, trade quality, and other backtest quality factors.

Use Monte Carlo as a filter, not as proof that a strategy will work live.