Monte Carlo

Interpreting Monte Carlo Results

How to read confidence bands, probability of ruin and wipeout, drawdown distributions, VaR, and what the results mean for strategy robustness.

Interpreting Monte Carlo Results

After a Monte Carlo simulation runs, the results appear in a dedicated tab alongside your backtest results. This article walks through every component and explains what to look for.

The Confidence Band Chart

The chart is the centerpiece of the Monte Carlo results. It shows the range of equity paths your strategy could have produced if the same trades had arrived in different orders.

Reading the Bands

The outer band (p5 to p95) covers 90% of all simulated paths. At any point along the horizontal axis, 90% of simulations had equity between the top and bottom edges of this band. Only 5% of paths went above the top edge, and 5% went below the bottom.

The inner band (p25 to p75) covers the middle 50% of paths. This is the range of typical outcomes. If you were to trade this strategy many times with different trade sequences, your experience would most likely fall somewhere within this band.

The median line (p50) is the middle path -- half of all simulations did better, half did worse. This is the middle-of-the-road equity trajectory for the strategy.

The actual path is your real backtest result, overlaid on the bands. Where it falls relative to the bands tells you how lucky or unlucky the specific trade sequence in your backtest was.

The bands always fan out
Confidence bands start narrow (all simulations begin at the same starting equity) and widen over time. This fanning effect is natural -- with more trades, there are more possible orderings, and the range of possible equity paths grows. Wider bands after many trades do not mean the strategy is getting worse. They mean uncertainty compounds, which is true for any strategy.

What the Bands Tell You

Narrow bands mean the strategy's outcome is not very sensitive to trade ordering. Whether the wins come first or last, the equity path stays in a similar range. This is a sign of robustness -- the result is not an artifact of a fortunate sequence.

Wide bands mean the strategy's outcome is highly path-dependent. Different trade orderings produce dramatically different experiences. Wide bands are typical of strategies with a few large wins and many small losses, or strategies with very high volatility. The result you saw in your backtest may not be representative of what you would actually experience.

If the bottom of the outer band (p5) dips below your starting equity, at least 5% of simulations lost money. If it dips significantly below, even a profitable strategy has a meaningful chance of producing a negative experience depending on trade sequence.

Probability of Ruin

This is the percentage of simulations where equity fell below the ruin threshold at any point during the simulation. The default threshold is 50% -- meaning the strategy lost half its starting capital.

A probability of ruin of 3% means that in 3 out of every 100 simulated trade sequences, the strategy's equity dropped below half of the starting capital at some point. The strategy may have recovered afterward, but the drawdown was severe enough to cross the ruin line.

What the Numbers Mean

0% -- No ruin in any simulation
No reshuffled sequence produced a drawdown severe enough to cross the threshold. This is ideal but only meaningful with a large number of simulations (1,000+).
Under 5%
Low risk. A small fraction of possible trade orderings would produce a catastrophic drawdown. The strategy is reasonably robust to sequence risk.
5-15%
Moderate risk. The strategy can survive most trade orderings but has a non-trivial chance of hitting severe drawdowns. Consider whether you could tolerate the ruin scenario.
Above 15%
High risk. A significant fraction of possible paths lead to severe drawdowns. The strategy may be profitable on average but is fragile to trade ordering. With bad luck in the sequence, you could face losses that are difficult to recover from.
Ruin probability is path-sensitive, not return-sensitive
A strategy can have positive expected returns and still have a high probability of ruin. If the strategy produces a few very large wins and many small losses, a run of losses at the start (before the big wins arrive) can push equity below the ruin threshold. The overall result is profitable, but the path to get there can be devastating depending on sequencing.
Why does the p5 band stay above the ruin threshold when Probability of Ruin is non-zero?
Because the confidence bands and Probability of Ruin measure different things. The p5 line is a snapshot at each point in time: at each trade index, it shows the 5th percentile of equity across all simulations at that moment. Different simulations can be in the bottom 5% at different trade indices -- the line is not tracking one fixed set of paths. Probability of Ruin tracks each individual path from start to finish: it asks whether each simulation's equity ever dropped below the ruin threshold at any point during its entire sequence. A simulation can dip below the threshold at trade 150, then recover to well above it by trade 500. That simulation counts toward ruin (because it touched the threshold), but it contributes a healthy equity value at later trade indices, pulling the p5 band up.

Probability of Wipeout

This is the percentage of simulations where equity reached effectively zero -- a complete loss of capital. This is more extreme than ruin and is usually near zero for any reasonable strategy. A non-zero wipeout probability is a serious red flag, especially when combined with leverage, which amplifies losses and makes wipeout more likely.

The engine treats equity as wiped out when it drops below 0.00005 -- effectively zero. This threshold exists because of floating-point precision: after a long series of compounding losses, equity can become an extremely small positive number that would display as 0.0000 when rounded. Rather than continuing to simulate trades on a functionally worthless account, the engine treats it as a full wipeout.

Once a wipeout is triggered, the simulation is permanently shut down. Equity is set to zero for all remaining trades and no recovery is possible within that simulation. This is checked after every single trade.

How is wipeout different from ruin?
Ruin means equity dropped below a threshold (default 50%) at any point but may have recovered afterward -- the strategy survived a crisis. Wipeout means equity reached effectively zero and the simulation stopped permanently -- the account is gone. A strategy can have a high probability of ruin (many paths dip below 50%) and a zero wipeout rate (none of those paths actually hit zero).

Max Drawdown Distribution

Your backtest produced one specific max drawdown. Monte Carlo shows you the distribution of max drawdowns across all simulated paths. This answers the question: was your backtest's drawdown typical, lucky, or unlucky?

The summary card shows the average max drawdown across all simulations and the 5th percentile (the worst-case drawdown that 95% of simulations stayed above). Pay particular attention to the 5th percentile -- this is a reasonable estimate of the worst drawdown you should prepare for.

Why is the Monte Carlo max drawdown usually worse than my backtest's?
Your backtest showed one trade sequence. Monte Carlo tests 1,000. Some of those sequences will cluster losses together more than your backtest did, producing deeper drawdowns. The Monte Carlo distribution is a more complete picture of the drawdown risk than the single number from your backtest.

Average Drawdown Metrics

Beyond the maximum drawdown, Monte Carlo also tracks the average depth and duration of all drawdown periods across simulations.

Average drawdown depth is the mean depth of all drawdown periods (only counting drawdowns that recovered). This tells you how deep the typical pullback is -- not the worst case, but the everyday experience of trading the strategy.

Average drawdown duration is measured in trades (not time). It tells you how many trades it typically takes to recover from a drawdown and reach a new equity high. Longer durations mean more patience required during losing streaks.

Drawdown Risk

The drawdown risk metrics provide a broader view of the underwater experience.

Time underwater
The percentage of time equity spent below its previous high-water mark, averaged across simulations. A strategy with 60% time underwater means you would typically be in a drawdown more often than not. Even profitable strategies can spend a surprising amount of time underwater.
Expected drawdown
The mean depth of all observed drawdowns across all simulations. This is the drawdown you should consider "normal" for this strategy.
Drawdown percentiles (p50, p75, p95)
The 50th, 75th, and 95th percentile drawdown depths. The p95 drawdown is especially important -- it tells you the drawdown that only 5% of simulations exceeded. This is a practical worst-case planning number.
When these outputs are informative
The two sections below read a spread of final returns, which only exists when the endpoint can vary. Under the default reshuffle method at 1x leverage, every simulation lands on the same final return (see How It Works), so Value at Risk and the confidence interval collapse to a single value -- they become informative with the Resample method, or when leverage makes trade order affect the endpoint.

Value at Risk (5%)

The Value at Risk (VaR) at the 5% level shows the final return at the 5th percentile of all simulations. In plain terms: 95% of simulated trade sequences ended with a return above this number. Only 5% ended worse.

If the VaR is positive, even the worst 5% of paths were profitable. If it is negative, there is at least a 5% chance the strategy would have lost money with a different trade sequence.

95% Confidence Interval

The confidence interval shows the range of final returns between the 2.5th and 97.5th percentiles. 95% of all simulated sequences ended within this range. It gives you a practical bracket for what the strategy's final result could be.

A narrow confidence interval means the final return is not very sensitive to trade ordering -- you would end up in a similar place regardless of sequence. A wide interval means the final result is highly path-dependent, and the specific return from your backtest could have been significantly better or worse with different trade ordering.

Leverage Effects

If your backtest used multiple leverage levels, Monte Carlo results are computed separately for each level. You can switch between them to see how leverage amplifies path risk.

At higher leverage, the confidence bands widen, the probability of ruin increases, and the drawdown distribution shifts deeper. A strategy that looks safe at 1x leverage can show alarming ruin probabilities at 3x or 5x. This is one of the most practical uses of Monte Carlo -- testing whether a leveraged strategy can survive the full range of possible trade sequences.

Putting It All Together

When reviewing Monte Carlo results, work through these questions in order:

1. Is the probability of ruin acceptable? If it is above 10-15%, the strategy has a meaningful chance of producing catastrophic losses regardless of its average return. This should be a hard constraint.

2. Where does the actual path fall in the bands? If your backtest result sits near the top of the confidence band, you may have been lucky. If it sits near the median, the result is typical and more trustworthy as a forward-looking estimate.

3. Can you tolerate the p95 drawdown? The worst 5% of simulations represent the bad-luck scenario. If the p95 drawdown would cause you to abandon the strategy or suffer unacceptable losses, the strategy is not suitable for your risk tolerance, even if the average case looks good.

4. How wide are the confidence bands? Wide bands mean high uncertainty. If the difference between the 5th and 95th percentile final return is very large, the strategy's outcome is a coin flip based on trade ordering. Narrow bands mean the result is robust regardless of sequence.

Monte Carlo is one piece of the puzzle
Monte Carlo tells you about path robustness -- can the strategy survive different trade orderings? It does not tell you about temporal robustness -- will the strategy produce similar trades on future data? Walk-forward analysis answers that question. Together, they give you a much more complete picture of strategy reliability than either alone.

Where to Go Next

How It Works -- the trade reshuffling method and simulation process.

Max Drawdown -- the single-backtest metric that Monte Carlo expands into a distribution.

Value at Risk -- the VaR metric in more detail.

Conditional Value at Risk -- expected loss beyond the VaR threshold.

Leverage & Position Sizing -- how leverage amplifies the risks Monte Carlo reveals.

Walk-Forward Analysis -- the complementary validation method for temporal robustness.