Risk of ruin math for crypto traders — why bet sizing decides survival

Edge is the headline metric. Sizing is the metric that decides whether you survive long enough to collect the edge. This article works through the math of risk-of-ruin, applies it to realistic crypto strategy parameters, and shows why most retail traders are mathematically guaranteed to be wiped out — not because their strategies are bad, but because their position sizing is set without regard to the underlying probability arithmetic.

What "risk of ruin" actually means

Risk of ruin is the probability that, over a sequence of trades, your account drops to zero or to a hard "ruin threshold" you've defined (say, -50% drawdown that you wouldn't psychologically recover from). It's a direct function of three inputs:

1. Win rate (p) — the fraction of trades that close in profit.
2. Reward-to-risk (R) — the average win divided by the average loss, in account-percentage terms.
3. Bet fraction (f) — the fraction of the account risked per trade.

For a strategy with positive expected value (i.e., one that should make money over many trades), risk of ruin is determined almost entirely by bet fraction. You can have an excellent edge and still go to zero with high probability if you size aggressively enough. The math doesn't care that the strategy is profitable on average. It cares about the path.

The simple version

For a strategy where every win pays the same R times the loss size (e.g., 1:2 risk-reward, fixed), and bet fraction is constant per trade as a fraction of current account, the classic risk-of-ruin formula is:

RoR = ((1 - edge) / (1 + edge)) ^ (units_to_ruin)

where:
  edge          = expected return per dollar risked
                = p * R - (1 - p)
  units_to_ruin = fraction of account that must be lost to ruin / bet fraction

This is the Kelly-derived formula and applies under simplifying assumptions: independent trades, constant edge, constant bet fraction in dollar terms (not percentage of account). The percentage-of-account version is slightly different but the intuition is identical.

Some worked examples to ground the intuition:

Two things stand out from the table:

1. For strategies with real edge (rows 1, 3), reasonable sizing (2% per trade) keeps risk of ruin very low — single-digit percent.
2. For the same strategies, sizing up to 10% per trade pushes risk of ruin to 28-50%, even with edge intact. The strategy didn't change. The size did. The math says half of these accounts die.

Kelly criterion, briefly

Kelly criterion gives the bet fraction that maximizes long-term geometric growth — equivalent to maximizing the rate of compounding. For a binary bet:

f* = p - (1 - p) / R

where:
  f* = optimal bet fraction
  p  = probability of winning
  R  = win/loss ratio

For a 65% win rate at 1:1 R:R, f* = 0.65 - 0.35 = 0.30. Thirty percent per trade. That's the "growth-optimal" size.

It's also the size at which your account equity behaves like a casino slot machine — extreme volatility, large drawdowns, and a feels-bad experience even when the long-run growth is theoretically optimal. Most professional traders use "fractional Kelly" — typically a quarter to a half of the full Kelly bet — to reduce variance at a small cost to expected growth. Quarter-Kelly on the above example is 7.5% per trade. Even that produces uncomfortable drawdowns.

The lesson is not "Kelly is too much." The lesson is that the size which is mathematically optimal for growth is dramatically larger than the size that's psychologically tolerable, which is dramatically larger than the size that retail traders end up using on bad strategies. Kelly is a useful upper bound. Real-world trading sizes should sit well below it.

Where retail goes wrong

The systematic failure pattern in retail crypto trading is:

1. Overestimate edge. They think their strategy has 60-70% win rate when it actually has 50-55%, or they think their R:R is 2:1 when slippage and fees bring it down to 1.3:1. So their perceived Kelly fraction is much higher than the real one.
2. Size near or above perceived Kelly. They think they should bet 25% per trade because their (overestimated) edge supports it. The real Kelly might be -2% (i.e., they shouldn't trade at all), and they're betting 25%.
3. Use leverage on top. A 25% notional bet at 5× leverage is a 125% effective bet. Mathematically impossible to size this way without ruin in any reasonable number of trades.
4. Don't measure. They don't track their actual win rate and average R:R rigorously. Without accurate inputs, the Kelly calculation is theatre. The right size is unknowable to them.

The result is a population of retail traders who size as if they have alpha they don't have. The math then does what the math always does, which is wipe most of them out within 3-12 months. This is observable in any retail-broker statistic; the percentages are roughly 70-80% of accounts going inactive within a year.

The crypto-specific factors

Crypto adds three structural problems to the standard risk-of-ruin picture:

1. Tail volatility. Equity strategies have ~20% intra-year vol; crypto has 60-100%+. The same Kelly fraction applied to crypto produces dramatically larger swings and drawdowns. You need smaller bet fractions to match the same psychological / financial pain tolerance.
2. Liquidation risk. Leveraged crypto positions get force-closed by exchange margin engines. A "stop loss" at -5% on a 5× leveraged position means a 1% adverse move triggers liquidation, often with worse fills than a market-stop would have given. This converts a soft stop into a hard zero on the position.
3. Correlated tails. Crypto positions are not independent. A bad market regime hits multiple positions at once. The risk-of-ruin math assumes independence; the reality has correlation that can spike to 0.8-0.9 during crashes. Multiple "uncorrelated" trades all lose together. Your effective bet fraction is the sum of correlated positions, not the per-position size.

For these reasons, professional crypto traders typically size at a quarter to a tenth of the Kelly fraction calculated from naive backtests. Effective bet fractions in the 0.5-3% range are normal for trading firms. Above 5% is aggressive. Above 10% is degenerate gambling regardless of the perceived edge.

Drawdown distribution

Even if you don't go to zero, you'll see drawdowns. The distribution of maximum drawdown over N trades, for a positive-edge strategy with bet fraction f and edge e, has a rough rule of thumb:

Expected max DD ≈ f × √(N / e × variance term)

The exact formula isn't easy to write closed-form (depends on path distribution, autocorrelation, sizing rules), but the scaling holds: doubling f doubles your expected max drawdown; quadrupling N (more trades) increases expected max DD by 2x; halving the edge increases expected max DD substantially.

Practical implication: if you size at 5% per trade and run 200 trades over a year on a strategy with 10% per-trade edge, expected max drawdown is in the 25-40% range even if everything goes well. You should be psychologically and financially prepared for a -30% drawdown on the way to a +60% year. If you can't, you're sizing too large for your risk tolerance regardless of what the strategy "should" return.

The right way to size

A practical sizing recipe:

1. Estimate edge honestly. Compute win rate and average R:R from a path-dependent backtest with realistic slippage and fees. Knock down the win rate by 3-5% as a reality margin. Real live performance is almost always worse than backtest because you can't fully model adverse selection.
2. Compute Kelly. Use the deflated edge estimate. Don't use leverage adjustments yet.
3. Take a fraction. Quarter-Kelly is the most common starting point. Eighth-Kelly is conservative. Half-Kelly is aggressive.
4. Cap at portfolio level. If your positions are correlated (most crypto longs are), the relevant Kelly is the portfolio Kelly with correlation accounted for. The simplest version: cap total exposure across all positions at ~Kelly_per_position × number_of_independent_positions, where "independent" usually means no more than 2-3 in crypto regardless of how many tickers you hold.
5. Stress test. Simulate 10,000 paths through the strategy with random outcome ordering. Look at the 5th-percentile drawdown across paths. If that's outside what you can stomach, reduce size further.
6. Cap leverage. Effective leverage above 3-5× is rarely justified by edge math. Anything above 10× is not a strategy choice; it's a path to liquidation given enough trades.

This recipe will produce sizes that feel "too small" relative to what most retail-trader content suggests. That's correct. Retail content optimizes for engagement; the math optimizes for survival.

The honest conclusion

Most retail crypto traders fail not because their strategies are bad — though some are — but because their sizing is calibrated to their hopes rather than the underlying probabilities. Risk-of-ruin math is unforgiving. It doesn't care about the trader's emotional state, recent winning streak, or conviction in the next trade. It compounds risk just as relentlessly as it compounds returns, and at the wrong size, the compounding goes one way.

The actionable takeaway is to size based on the math, not on the feeling. Compute Kelly from honest inputs, take a fraction of it, cap exposure across correlated positions, and don't increase size after winning streaks. None of this is exotic. All of it is non-obvious to traders who haven't done the arithmetic. The arithmetic is the difference between trading for ten years and being out of the game in eighteen months.

If you want to compute your own risk of ruin, the formulas above are enough to set up a spreadsheet in 30 minutes. Plug in your honest win rate and R:R. Try different bet fractions. Look at the resulting RoR numbers. If any of them are above 5-10% and represent how you actually trade, you have a problem that no amount of "better signals" will fix. The fix is a smaller number on the bet-size cell. Everything else is downstream of that.