Why positive expectancy is not enough
A coin that lands heads 55% of the time has positive expectancy. Bet everything on every flip, and the first string of tails ends the game before the edge compounds into anything meaningful. The math of ruin is entirely separate from the math of edge.
Most new traders understand this in theory. In practice, they size like the edge is guaranteed to manifest in the next 20 trades. It is not. The correct question is not "does this strategy have edge?" It is "given my position size, what is the probability that the natural variance of this edge destroys my account before I can realize the returns?"
That probability has a name: risk of ruin.
The formula
For a system where you risk a fixed dollar amount L on each trade, and your win probability is p:
Risk of Ruin — fixed dollar risk per trade
R = ((1 − p) / p) ^ (C ÷ L)
R = ruin probability (0 to 1)
p = win probability (e.g. 0.55 for 55%)
C = total account size in dollars
L = fixed dollar risk per trade
Two worked examples at opposite ends:
Example A — conservative sizing
Win rate: 55%, Account: $10,000, Risk per trade: $100 (1%)
R = ((0.45 / 0.55) ^ 100) = 0.818 ^ 100 ≈ 0.00001 (0.001%)
Ruin is effectively impossible with correct sizing.
Example B — aggressive sizing
Win rate: 55%, Account: $10,000, Risk per trade: $500 (5%)
R = ((0.45 / 0.55) ^ 20) = 0.818 ^ 20 ≈ 0.019 (1.9%)
Same edge, same account. ~1 in 50 chance of blowing up.
The only thing that changed between examples A and B is the position size. Same strategy, same win rate, same account. One is essentially safe; the other will destroy 1 in 50 traders before the edge compounds.
Now run 1,000 traders on strategy B. Roughly 19 of them blow up entirely. They conclude the strategy does not work. They are wrong. The strategy works fine. They sized incorrectly.
Why the simple formula underestimates real ruin risk
The formula above assumes every trade risks exactly the same dollar amount. Real trading does not work that way. Several factors make ruin more probable than the simple calculation suggests.
Variable payoffs
Most real strategies have non-uniform returns. Losses are not always 1R. Slippage on stop-outs, gap events, and partial fills all mean some losses are 1.5R or 2R. The formula treats every loss identically, which underestimates tail risk.
Correlated losses
The formula assumes independent trades. If you trade the same asset in the same session repeatedly, losing trades cluster during unfavorable market conditions. A 6-trade losing streak that the math calls unlikely becomes routine when all 6 trades occur in the same choppy Thursday session.
Drawdown-based sizing
Many traders increase size after wins and reduce it after losses, but the calculation assumes fixed sizing throughout. If you size up after a run of wins and hit a reversal, actual losses exceed the formula's prediction.
Psychological ruin threshold
The formula assumes you stop at zero. In practice, many traders stop much earlier — when the drawdown becomes emotionally unbearable and they deviate from the system. Effective ruin for most people is a 40–50% drawdown, not 100%. This makes the true stopping probability significantly higher than the formula implies.
For any strategy you intend to trade at size, use Monte Carlo simulation instead of the closed-form formula. Resample your trade log, run 10,000 scenarios, and read the distribution of outcomes directly.
The Kelly connection
The Kelly criterion gives you the mathematically optimal fraction of your account to risk on each trade to maximize long-run growth. Sizing above Kelly is provably suboptimal — it produces lower long-run returns than Kelly, while simultaneously increasing ruin probability.
Kelly fraction — win/loss payoff version
f* = (p × b − (1 − p)) ÷ b
f* = fraction of account to risk per trade
p = win probability (e.g. 0.55)
b = average win ÷ average loss (payoff ratio)
Example: 55% win rate, 1.5:1 payoff ratio
f* = (0.55 × 1.5 − 0.45) ÷ 1.5 = (0.825 − 0.45) ÷ 1.5
f* = 0.375 ÷ 1.5 = 0.25 (25% of account per trade)
Full Kelly says risk 25%. In practice, this produces vicious drawdowns.
Full Kelly almost always produces larger drawdowns than traders can tolerate emotionally, even though the math is correct. The standard professional answer is half-Kelly: take the Kelly fraction, divide by two, and trade that. Half-Kelly produces roughly 75% of the long-run growth of full Kelly while cutting drawdown variance roughly in half.
For a strategy with realistic estimation error (which is all strategies in practice), half-Kelly is usually the aggressive end of reasonable. Most systematic traders target a fraction between one-quarter and one-half Kelly.
What ruin probability looks like in practice
The table below shows how ruin probability changes with position size for a 55% win rate system with a 1:1 payoff ratio on a $10,000 account.
Risk of ruin vs position size — 55% win rate, 1:1 RR, $10k account
At 10% risk per trade, one in seven traders with this exact system blows up before the edge has time to accumulate. They are not doing anything wrong with the strategy. They are just expressing a position size that the natural variance of the system cannot sustain.
The rule of thumb that emerges: risk 1–2% per trade on a system with typical variance, and you can trade through a 30-trade losing streak without account destruction. That streak will feel catastrophic emotionally. Mathematically, it is routine variance on a 55% system.
How to set your position size
The right position size is not the one that maximizes return on a good month. It is the one that keeps ruin probability low enough that you survive to a large enough sample to realize your edge.
A practical workflow:
1. Estimate your system's parameters from a real trade log
Win rate and average payoff ratio from at least 100 trades. Not from backtesting alone — backtesting overstates win rate and underestimates average loss. Use walk-forward out-of-sample results or live paper trading data.
2. Compute your Kelly fraction
Use the formula above. Then take half. That is your upper bound on risk per trade.
3. Run a Monte Carlo simulation
Resample your trade log 10,000 times at your proposed position size. Check what fraction of simulations reach your ruin threshold (typically -50% drawdown). Target below 1%. If the fraction is higher, reduce position size until it is not.
4. Start smaller than the model says
Your parameter estimates have estimation error. The real win rate is not exactly the backtest win rate. Build in a safety margin by starting at one-quarter to one-half of your calculated optimal size, and scale up only after live results confirm the edge.
5. Set a daily or weekly loss limit
Risk of ruin assumes you keep trading through every losing streak. Most people do not — they escalate, revenge trade, or abandon the system entirely. A hard daily loss limit (typically 3–5% of account) interrupts runaway losing sessions before the psychological damage compounds the mathematical damage.
Kelly criterion and position sizing →
For a detailed walkthrough of the Kelly formula, fractional Kelly sizing, and worked examples across different strategy types, read the Kelly Criterion Position Sizing guide.
Summary
- Risk of ruin is the probability that natural variance destroys your account before your edge compounds — positive expectancy does not make ruin impossible
- The closed-form formula R = ((1−p)/p)^(C/L) gives a useful approximation; Monte Carlo simulation on your actual trade log is more accurate
- The formula underestimates real ruin because it assumes independent, fixed-size trades — correlated losses, slippage, and psychological stopping points all make actual ruin more likely
- Sizing above full Kelly produces lower long-run returns and higher ruin probability; half-Kelly or less is the professional standard
- At 1–2% risk per trade, ruin probability on typical strategies is negligible across thousands of trades
- Start smaller than your model suggests, confirm the edge is live, and scale up only after real sample validation
Frequently asked questions
What is risk of ruin in trading?
Risk of ruin is the probability that a sequence of consecutive losses draws your account down to zero (or to a defined stopping threshold) before you recover. A strategy with positive expectancy can still have a non-trivial risk of ruin if position sizes are too large relative to the account size and the variance of the system.
How do you calculate risk of ruin?
The simplified formula for fixed-risk trading is: R = ((1 − p) / p) ^ (C / L), where p is your win probability (expressed as a decimal), C is your account size, and L is the fixed dollar amount you risk per trade. This formula assumes each trade risks the same dollar amount. In practice, most traders use Monte Carlo simulation because real strategies have non-uniform trade returns.
What is an acceptable risk of ruin?
Professional traders typically target a risk of ruin below 1%, and many sophisticated operators target below 0.1%. Anything above 5% means a meaningful chance of losing the account before the edge manifests. Bringing ruin probability below 0.5% usually requires risking no more than 1–2% of capital per trade on systems with typical win rates.
Does a high win rate eliminate risk of ruin?
No. A high win rate reduces ruin probability but does not eliminate it. A system with a 65% win rate and 5% risk per trade still carries meaningful ruin risk across hundreds of trades. The win rate, payoff ratio, position size, and account size all interact. Simulating your specific parameters is the only reliable way to estimate your actual ruin probability.
How does the Kelly criterion relate to risk of ruin?
Kelly sizing is the position size that maximizes long-run account growth. Sizing above Kelly produces lower long-run growth than Kelly, and dramatically increases ruin probability. Sizing at half-Kelly (the standard professional recommendation) produces roughly 75% of Kelly growth while keeping drawdowns and ruin probability at much more manageable levels.
Validate your strategy before sizing up
Before scaling a strategy, verify that the edge you are sizing around is real and not curve-fitted. Run our free Strategy Overfitting Score to quantify how likely your backtest is fitting noise — then book a diagnostic call to review your sizing approach.