Home › Documentation › Validation & robustness
Analysis 4

Monte Carlo Simulation

Available in the Monte Carlo tab of the Strategy Performance Report. It answers two questions: what range of results is it reasonable to expect from this strategy? and was the drawdown I suffered lucky or typical? To do so it generates 1,000 alternative histories from your real trades and shows you where yours falls.

Two methods, chosen at the top

Bootstrap with replacement (default)

The standard method, the same one AmiBroker uses. Each simulation draws as many trades as you have, from the real ones and with repetition: a good trade may come out twice and a bad one not at all, or the other way round. Both the path and the final result change, so you get a realistic range of what the strategy could have delivered.

Random order

Shuffles the same trades into a different order. The final result does not change (the sum does not depend on the order), only the path and, with it, the drawdown. It only answers «was I lucky with the order in which the losses came?».

The program remembers the chosen method. In both, the real sequence counts as one more simulation, so the percentage never reaches a misleading 100 %.

What it shows
  • → The real equity curve (amber line)
  • → The p25–p75 band of the simulations
  • → The p5–p95 band (the possible paths)
  • → The median of the 1,000 simulations (blue line)
  • →Real drawdown vs worst case (p95)
  • →With bootstrap, the final result at the worst 5 %, the median and the best 5 %
  • →Coloured rating: % of simulations with a worse drawdown than the real one
How to read it
50 % is what pure chance would give. The further below, the more clustered the losses were; the further above, the better spread.
80–100% → SOLID
Your real drawdown is smaller than that of the vast majority of simulations. The curve holds up well even if the sequence of trades changes.
60–79% → GOOD
The losses are better spread than in a typical history.
40–59% → ACCEPTABLE
Typical behaviour, what chance would give. Look at the worst case (p95) to size the real risk.
20–39% → WATCH
The losses came somewhat more clustered than normal. Expect worse streaks than in a typical history.
0–19% → WARNING SIGN
Your real drawdown is among the worst of all the simulations: the losses came very clustered in time. It usually means the strategy depends on one particular market regime.

A real example

An @ES strategy with 235 trades made $288,675. With bootstrap, in the worst 5 % of histories it would have ended at $159,084, at $288,592 in the median and at $423,443 in the best 5 %. Its real drawdown ($49,043) beats that of 28 % of the simulations: WATCH zone, with a p95 worst case of $68,323.

That $159,084 is the prudent figure to get an idea of what can go wrong, far more useful than the single result of the backtest.

Performance

With 1,000 simulations and hundreds of trades, the full calculation (draws + equity + drawdown + percentiles) takes a fraction of a second and runs in the window itself, with no waiting.

1,000
simulations
2
methods

The simulations are reproducible: when you come back to the tab you will see exactly the same chart.

Try it yourself

AniQuant can be tried free for 14 days, with every module and no card. If you have the book, 30.

← Previous
Parameter robustness test
Next →
Market regime: is the market on the strategy's ground?
More in Validation & robustness