Backtestify
Sign in
All articles
Risk of RuinSeptember 26, 202610 min read

95th Percentile Simulations to Cut Your Risk of Ruin

Simulation first guide to estimate and control risk of ruin with Monte Carlo and bootstrap percentiles (95th, 99th), concrete sizing rules, and evidence...

!Isometric illustration of risk barriers and equity paths

Risk of ruin is the probability that your equity crosses a failure boundary you define in advance, and the immediate move is to simulate your strategy with a Monte Carlo or bootstrap model and size positions with fractional Kelly or a fixed fraction of capital. The number depends most on your risk per trade, your win rate, and how dispersed your winners and losers are. Platforms like backtesting tools exist because that estimate needs real trade data, not guesswork.

TL;DR:

  • The risk of ruin depends on your risk per trade, win rate, and how dispersed your wins and losses are, making accurate simulation essential.
  • Classical formulas are only useful for intuition because they rely on unrealistic assumptions like equal bet sizes and independent outcomes.
  • Monte Carlo and bootstrap methods provide more realistic ruin estimates by resampling actual trade data and accounting for path dependence and costs.
  • Setting conservative thresholds at the 95th or 99th percentile in simulations helps prevent underestimating the probability of large, adverse drawdowns.
  • Regularly re-running simulations after significant equity swings and maintaining disciplined sizing measures reduce the likelihood of unexpected ruin.

Table of Contents

What risk of ruin actually measures versus drawdown

Risk of ruin is the probability of breaching a barrier you set ahead of time, and that barrier rarely means losing every dollar. According to Investopedia, it is the chance that losses grow large enough that recovery becomes impossible, which can mean hitting a margin call, a prop firm's daily loss limit, or simply a floor you personally will not trade below.

Drawdown, by contrast, is a single observed decline from a peak. Ruin is a probability calculated across many possible futures.

Barriers traders actually use include:

  • A static floor, such as losing 50% of starting capital.
  • A peak-relative drawdown, such as a 20% pullback from your equity high.
  • A broker margin call triggered by a maintenance level.
  • A prop-firm rule, like a fixed daily or overall loss cap.

That asymmetry is why ruin thresholds matter more than raw drawdown numbers.

How to calculate risk of ruin: classical formulas and their limits

The oldest version comes from gambler's-ruin math: with equal-size bets, a fixed edge, and independent outcomes, there is a closed-form probability of hitting zero before reaching a target. A Kelly-equivalent shortcut often cited alongside it approximates ruin probability as roughly e to the power of negative twice your bankroll divided by your average bet size in Kelly units, shrinking fast as your effective bankroll multiple grows.

These formulas are useful for intuition, not for trading decisions, because they rest on assumptions trading rarely satisfies.

  1. Equal bet size across every trade, which ignores position sizing that scales with conviction or volatility.
  2. Independent and identically distributed outcomes, meaning no streaks, no regime shifts, and no correlation between trades.
  3. No transaction costs, spreads, slippage, or financing charges factored into the payoff.

Real trading breaks all three. R-multiples vary trade to trade, returns show fat tails that classical models understate, and path dependence means the order of wins and losses changes the outcome even when the average stays the same. The Stanford Kelly primer notes that naive application of Kelly-style edge estimates, without accounting for estimation error, often leads to ruin despite the formula's long-run optimality. Closed-form ruin math is a starting intuition, never a governance tool on its own.

Monte Carlo and bootstrap methods for realistic estimates

Because closed-form formulas assume conditions trading does not have, practitioners turn to resampling. The method is straightforward: take your historical trades, resample them thousands of times (or simulate returns from a fitted distribution), apply your actual sizing rule to each resampled sequence, and record how often the equity path crosses your chosen barrier.

The output is a distribution, not a single number. The median outcome tells you what a typical path looks like, but policy decisions should reference the tail. Research on Monte Carlo for trading systems recommends setting governance thresholds at the 95th or 99th percentile of the bootstrap distribution rather than relying on the single in-sample maximum drawdown, which understates plausible adverse outcomes.

Inputs that materially move the result:

  • Horizon length, since longer simulated periods surface deeper drawdowns.
  • Sizing rule, because fixed-fraction and fixed-dollar sizing produce different path dependence.
  • Slippage and cost assumptions, which compound over thousands of simulated trades.
  • Tail-stress scenarios, like widening the historical loss distribution to test fragility.

Pro Tip: Run the simulation twice, once with historical costs and once with costs doubled, to see how sensitive your ruin estimate is to execution quality.

Choosing an acceptable risk of ruin threshold

There is no universal correct number, only a threshold that matches your capital's purpose and your tolerance for variance.

Once you pick a threshold, work backward into rules you can actually follow:

  • Translate the target percentile into a maximum risk per trade, since that single input drives most of the ruin estimate.
  • Set a maximum drawdown rule that matches the peak-relative barrier used in your simulation.
  • Re-run the simulation under pessimistic inputs, lower win rate, wider losses, higher costs, to confirm the threshold still holds.

A threshold that survives a stress test is worth more than one that only looks good on your best-case data.

Reducing your risk of ruin: sizing and monitoring in practice

Lowering risk of ruin usually comes down to sizing discipline and a plan for what happens when things go wrong.

  1. Use fractional Kelly, betting half or a quarter of the full Kelly stake, since the Berkeley analysis of the Kelly criterion shows full Kelly maximizes long-term log wealth but produces large short-term drawdowns that most traders cannot stomach.
  2. Cap risk per trade at a small, fixed percentage of capital, and treat overlapping or correlated positions as a single combined risk exposure rather than separate bets.
  3. Set a kill-switch threshold using your simulation's percentile output rather than a round number picked out of habit.
  4. Monitor live equity against the bootstrap percentile bands monthly, so a real drawdown that tracks the 95th percentile band triggers a review before it becomes a crisis.

Pro Tip: Re-run your Monte Carlo simulation after any large equity swing, win or loss, because the inputs that generated your original ruin estimate may no longer reflect your current edge.

Estimation error in your edge matters more than error in your variance assumptions, which is why conservative sizing beats precise-looking math built on a shaky win-rate estimate.

!Reducing your risk of ruin: sizing and monitoring in practice — overview diagram

A practical workflow for running your own simulation

A defensible risk of ruin estimate follows a repeatable process, not a one-off calculation.

  1. Build a data checklist: trade-by-trade entry, exit, size, and realized cost, including slippage and fees, not just theoretical fills.
  2. Choose simulation defaults: a bootstrap of several thousand resampled paths, a horizon long enough to cover your typical holding period many times over, and at least one stress scenario with widened losses.
  3. Apply your actual sizing rule to every resampled path so the ruin estimate reflects real path dependence, not an idealized flat bet.
  4. Read the 95th and 99th percentile outcomes and convert the more conservative one into a concrete kill-switch level or a reduced risk-per-trade rule.

This is the same sequencing traders' Second Brain's guide to risk-of-ruin math points to: the estimate is only actionable once the barrier, horizon, sizing rule, and costs are fully specified.

Why evidence-based backtesting matters for a trustworthy estimate

A ruin estimate is only as good as the trade data behind it. Some backtesting platforms run backtests against real historical market data, report metrics like win rate, profit factor, and maximum drawdown, and forecast future price movement with probability statistics rather than a single optimistic number. Publishing results for both winning and losing strategies, rather than only the winners, can help keep the underlying edge estimate honest, which matters because an inflated edge produces an artificially low ruin probability.

Folding risk of ruin into routine habits

I re-run percentile checks after sharp equity moves and keep sizing at a fraction of full Kelly rather than the full stake. The math only helps if it becomes a weekly habit, not a one-time report you file away.

— WAJDI

Try the simulation workflow yourself

Reading about Monte Carlo percentiles is one thing, running them against your own trades is another. Some platforms let you backtest a strategy against real historical data, then compare the original version against an improved one side by side, so you can see how a sizing change actually shifts your drawdown and ruin numbers instead of guessing.

Backtestify

Start with the step-by-step backtesting guide to prepare your trade data, then explore Backtestify directly. If you also want a separate log for tracking bet sizing and outcomes across accounts, Betlog is a partner tool built for that. Free tiers often let you start testing, and paid plans unlock unlimited backtesting, improvement, and forecasting; see the platform's website for current pricing details.

Sources

FAQ

What is the risk of ruin theory?

Risk of ruin theory models the probability that a trader's or gambler's capital falls to a predefined failure point before reaching a goal, based on bet size, edge, and payoff variability. It originated in gambling mathematics and was later adapted to trading, where the Investopedia definition frames it as the point where recovery becomes practically impossible rather than always meaning total loss.

What is an acceptable risk of ruin?

There is no single correct number, since it depends on your capital's purpose and time horizon.

What are the 4 types of risk?

Common risk categories used in trading and finance include market risk, credit risk, liquidity risk, and operational risk, though frameworks vary by source and context. Risk of ruin itself is not one of these categories, it is a probability measure that can be driven by any combination of them.

What is the difference between risk of ruin and drawdown?

Drawdown is a single observed decline from an equity peak, a fact about what already happened. Risk of ruin is a forward-looking probability, usually generated through Monte Carlo or bootstrap simulation, that estimates how often a defined failure barrier gets breached across many possible future paths, as described in research on Monte Carlo methods for trading systems.

Recommended

Read next

Prove Risk Adjusted Returns With a Sharpe and Sortino Worked Example

Continue

Want these checks applied to your own rules automatically?

Run a backtest