OptiNod Academy

Risk of Ruin: Position Sizing Determines Account Survival

Even with the same win rate and reward-to-risk ratio, risk of ruin rises nonlinearly as the percentage risked per trade increases. A strategy can have positive expectancy, but oversizing can push the account into an unrecoverable drawdown after a single losing streak.

> Risk of ruin is determined by three variables: win rate, reward-to-risk ratio, and bet size. *Even with positive expectancy, an oversized position* can drive an account into an unrecoverable drawdown after one losing streak.

Risk of ruin is the probability that even a trading system with positive long-term expectancy will lose enough capital to become unrecoverable. The concept comes from gambling theory, and the definition is simple: if you repeat the same win rate and reward-to-risk ratio indefinitely, it is the probability that capital reaches zero or a predefined lower boundary. Win rate, reward-to-risk ratio, and the percentage risked per trade together determine that probability.

The popular interpretation stops at expectancy. If win rate and reward-to-risk produce positive expectancy, traders assume the system will “make money eventually.” The standard expectancy formula is not wrong. But it assumes unlimited capital and unlimited trials. Real accounts are finite, and once the account reaches zero, the game ends. This is why traders can go broke even with a positive-expectancy system. The real problem is the percentage risked per trade.

This article unpacks three points. First, even with the same edge, risk of ruin does not increase linearly with bet size. It rises sharply and nonlinearly. Second, even at a 50% win rate, a 10-loss streak is not statistically rare; as the number of trades grows, it becomes almost inevitable at some point. Third, large drawdowns are not repaired by gains of the same size. A -50% drawdown requires a +100% return to recover. When these three forces combine, the variable that determines survival narrows down to one thing: sizing.

Positive Expectancy Does Not Prevent Ruin

Positive expectancy describes the average outcome over infinite repetitions. It does not guarantee that finite capital will avoid crossing zero before that average is realized. The gambler's ruin problem shows this structure clearly. If each trade wins or loses a fixed unit and the game ends when capital reaches zero, the probability of hitting zero can still be meaningful even when the win rate is above 50%, especially when starting capital is small.

Consider an even-money game where the probability of winning is greater than the probability of losing. With a 55% win rate and a 1:1 reward-to-risk ratio, the expected value of one trade is positive. When capital is expressed in loss units, the probability of ruin can be approximated by raising the ratio of losing probability to winning probability to the power of the number of capital units. If you risk 2% of capital per trade, the account has 50 units of capital. In that case, the probability of ruin falls to 0.004%. It is effectively negligible.

Now keep the same 55% win rate and change only the bet size. If you risk 10% per trade, the account has 10 capital units, and the probability of ruin jumps to 13.4%. If you risk 25%, the account has 4 capital units, and the probability of ruin reaches 44.8%. The edge is unchanged. Only the percentage risked per trade increased. Yet the probability of ruin moved from 0.004% to nearly a coin flip. Positive expectancy does not explain this difference. There is a clear zone where sizing alone can make ruin common, even with a positive edge.

The Same Edge Can Produce Nonlinear Ruin Risk

If risk of ruin increased gently in proportion to bet size, sizing would be a simple trade-off between return and stability. The actual curve does not behave that way. Because the structure raises the ratio of losing probability to winning probability to the power of capital units, risk of ruin accelerates as bet size increases. At small bet sizes, the curve stays almost flat near zero. After a certain point, it bends sharply upward.

For a 55% win rate at even money, risk of ruin by bet size makes the nonlinearity clear.

• 2% risk: 50 capital units, about 0.004% risk of ruin. Effectively zero.
• 5% risk: 20 capital units, about 1.8% risk of ruin. Still low.
• 10% risk: 10 capital units, about 13.4% risk of ruin. This is where the curve starts to bend.
• 20% risk: 5 capital units, about 36.7% risk of ruin.
• 25% risk: 4 capital units, about 44.8% risk of ruin. Almost one in two.

When the bet size increases two and a half times from 2% to 5%, risk of ruin rises from 0.004% to 1.8%. But when the bet size increases two and a half times from 10% to 25%, risk of ruin jumps from 13.4% to 44.8%, a 31 percentage-point increase. The multiplier is the same, but the increase in risk is not. Each step becomes more expensive as you move to the right side of the curve.

When the edge is thinner, the entire curve becomes steeper. If the win rate falls to 52%, the risk of ruin for a 5% bet rises from 1.8% to 20%, and for a 10% bet from 13.4% to 44.9%. The thinner the edge, the less room there is for sizing. The same percentage risk pushes the account toward ruin much faster when the edge is small.

A 10-Loss Streak Is Not Rare, Even at a 50% Win Rate

Underestimating losing streaks is the most common starting point for oversizing. If you look only at the loss probability of a single trade, a long losing streak feels remote. But for independent trades, the probability of consecutive losses is the loss probability raised to the number of losses in the streak. At a 50% win rate, the probability of one loss is 0.5. Three losses in a row is 0.5 cubed, or 12.5%. Five losses in a row is 3.125%. Ten losses in a row is 0.098%. Viewed as a single event, a 10-loss streak looks easy to ignore: one in 1,024.

The problem is that trading is not a single event. As the number of trades grows, the probability that this one-in-1,024 streak appears somewhere rises quickly. With a 50% win rate over 100 trades, the probability of at least one 5-loss streak is 81%, at least one 7-loss streak is 31.8%, and at least one 10-loss streak is 4.4%. Over 500 trades, the probability of experiencing at least one 10-loss streak rises to 21.5%. For an active trader, a 10-loss streak is something to expect eventually.

Raising the win rate to 55% does not create a large safety zone. At a 55% win rate over 100 trades, a 5-loss streak or worse still appears with 64.7% probability. Over 500 trades, a 10-loss streak or worse appears with 8.8% probability. The reduction in losing-streak risk from a 5 percentage-point higher win rate is smaller than most traders expect. If sizing is not based on the worst losing streak your win rate can produce, you are placing yourself where one statistically normal streak can end the account.

This is where sizing and losing streaks connect directly. If a trader risks 25% of capital per trade and hits a 10-loss streak, remaining capital is 0.75 to the 10th power, or 5.6%. The account is effectively finished. With the same 10-loss streak at 2% risk per trade, remaining capital is 81.7%, limiting the drawdown to 18.3%. The system determines the length of the losing streak. Bet size determines whether that streak destroys the account or stays within a tolerable drawdown.

Drawdowns Are Not Recovered by Equal-Sized Gains

The asymmetry of recovery makes risk of ruin even steeper. When capital falls by a given percentage, the same percentage gain does not return the account to breakeven. Returns are calculated on the reduced capital base. A -10% drawdown requires +11.1% to recover. A -20% drawdown requires +25%. A -30% drawdown requires +42.9%. Up to this point, the difference may not look large.

As the drawdown deepens, the required return rises sharply. A -50% drawdown requires +100%. A -75% drawdown requires +300%. A -90% drawdown requires +900%. A trader who loses 90% of capital must multiply the remaining capital by ten just to return to breakeven. With the same edge, that is close to impossible in practice. Because of this asymmetry, once the account falls into a deep drawdown, statistical recovery becomes increasingly distant even if the system still has positive expectancy.

A real market example shows the scale of this asymmetry. BTC peaked at $69,000 on November 10, 2021, then fell to $15,476 on November 21, 2022. That was a drawdown of about 78%. Recovering that 78% drop and returning to the prior high required an increase of roughly 346% from the low. BTC eventually recovered, but the recovery took time, and capital that was liquidated or stopped out during the 78% drawdown could not participate in it. An asset recovering and an account holding that asset recovering are different problems.

When this asymmetry combines with sizing, risk multiplies. If a trader risks 25% per trade and hits a 7-loss streak, only 13.3% of capital remains, and breakeven requires about +649%. With the same 7-loss streak at 5% risk per trade, 69.8% of capital remains, and +43% is enough to recover. Bet size determines the drawdown produced by a losing streak, and that drawdown asymmetrically determines the difficulty of recovery. Increasing size does not only increase the drawdown. It also reduces the probability of escaping it.

Kelly Criterion Is an Upper Limit to Stay Under

The Kelly Criterion gives one bet size that maximizes long-term capital growth. Given win probability p and reward-to-risk ratio b, the optimal fraction is calculated by subtracting the losing probability divided by b from p. With a 55% win rate at even money, the Kelly fraction is exactly 10%. At that size, expected log growth per trade is maximized. Popular understanding often stops here and treats the Kelly fraction as the number to use.

Using full Kelly maximizes long-term growth, but it comes with large drawdown volatility. An account trading at the Kelly fraction should expect drawdowns near half its value as part of normal behavior. In live trading, that drawdown is psychologically difficult to withstand, and even a small error in estimated win rate or reward-to-risk can push the actual bet size above Kelly. This is why practitioners use fractional Kelly, usually half Kelly or less. Half Kelly at 5% preserves roughly 75% of full Kelly's growth rate while sharply reducing drawdown volatility.

Once you exceed Kelly, the growth rate itself breaks down. In a 55% win-rate, even-money game, the expected log growth per trade is maximized at the Kelly fraction of 10%, at about 0.00501. At half Kelly, or 5%, it is 0.00375, preserving much of the growth. But at double Kelly, or 20%, log growth turns negative. At roughly twice Kelly, long-term growth crosses zero and becomes negative. Even though the system still has positive expectancy, capital declines over the long run once the bet size exceeds twice Kelly. The same structure appears in other combinations, such as a 2:1 reward-to-risk ratio with a 40% win rate: growth crosses zero near twice Kelly.

This becomes the sizing rule. Kelly is an upper limit, and live trading should stay at half Kelly or below. If you rely on positive expectancy and bet above Kelly, you are deliberately entering the zone where a positive edge turns into negative long-term capital growth.

Sizing Checklist: What to Confirm Before Trading

Decisions that reduce risk of ruin must be made before the trade begins. If you decide size after seeing an entry signal, you are already late. Fix the following items as trading rules.

• [ ] Per-trade risk limit: Set the maximum loss on one trade at 2% of capital or less. Calculate stop distance and position size together so that the actual loss at the stop does not exceed 2% of capital.
• [ ] Losing-streak scenario: Calculate in advance how much capital remains after a 10-loss streak at your system's loss probability. At 2% risk, 81.7% of capital remains after 10 losses, limiting the drawdown to 18.3%. Decide before entry whether you can tolerate that drawdown.
• [ ] Kelly ceiling check: Calculate the Kelly fraction from your win rate and reward-to-risk ratio, then confirm that actual risk does not exceed half of that value. With a 55% win rate at even money, Kelly is 10%, so actual risk should stay at 5% or below.
• [ ] Edge estimation error: Win rate and reward-to-risk are estimates and may be too optimistic. Subtract 3 to 5 percentage points from the estimated win rate, then recalculate Kelly and risk of ruin to check whether the size remains safe under a more conservative assumption.

When Sizing Rules Stop Working

The checklist above depends on one assumption: each trade is independent, and win rate and reward-to-risk remain stable. When that assumption breaks, the same size becomes more dangerous.

The first case is when trades are not independent. If you hold multiple positions in the same direction at the same time, they are effectively one larger bet. If you risk 2% each on BTC long, ETH long, and SOL long, you are risking 6% in one direction. If the market falls together, all three positions can hit their stops at the same time. For highly correlated positions, combine the risks and keep aggregate risk within the single-trade limit.

The second case is when reward-to-risk is overstated. The average reward-to-risk in a backtest falls after slippage and fees, and if a few large winners lift the average, the median may be much lower. If you calculate Kelly from the average alone, the resulting size will be too large, and the real position may sit above Kelly. Evaluate the full distribution of reward-to-risk, including the median.

The third case is when win rate changes over time. A trend-following system that works only in trending markets can see its win rate fall sharply in sideways regimes. If you apply sizing based on the trending-market win rate during a range-bound market, the bet may exceed Kelly for that regime. First check whether the current market environment is one where your system works. If not, reduce size or stop trading. Risk of ruin is determined by the losing streaks that accumulate while incorrect sizing is repeated, and the only variable you can control is the percentage risked per trade.