Skip to content

Kelly Criterion and Risk of Ruin: What the Sizing Math Assumes and Where It Misleads

Share
Smoked-glass tower pierces a horizontal glass pane, orange glow at the break point; headline: LIMIT BEFORE FORMULA

Every position starts with the same question: how much of the account goes into this trade? The Kelly criterion promises an exact answer. Feed it your win rate and the size of your average win against your average loss, and it returns a share of your account. No gut feeling. Just arithmetic.

That precision is the appeal. It is also the trap.

The same arithmetic is used wherever repeated bets are sized, but this piece stays with trading. It explains what the Kelly criterion calculates, what it quietly assumes, and why risk of ruin is the more useful question for anyone trading an account with a loss limit. It does not recommend a size, and it does not hand you a fraction to use.

Key Takeaways

  • The Kelly criterion turns a win rate and a payoff ratio into the share of the account that maximizes long-run growth under its assumptions.
  • Both inputs are estimates in trading, and an overestimated edge produces a position that is too large.
  • Sizing too large hurts growth more than sizing too small, so estimation errors do not cancel out.
  • The formula knows nothing about a hard loss limit; risk of ruin measures exactly that.
  • Fractional Kelly trades some growth for smaller swings, which is a choice about your limit, not a formula result.

What the Kelly Criterion Calculates

The Kelly criterion answers a narrow question. If you could repeat the same trade over and over, what share of your current account should you put at risk each time so that the account grows fastest over the long run?

It needs a pair of inputs. The first is your win rate, the share of trades that end as winners. The second is your payoff ratio, the average win divided by the average loss.

How the pair relates, and where the breakeven line between them sits, is covered in the breakeven win rate behind a ratio. This article builds on that and does not repeat it.

With the loss rate being whatever share of trades is not a win, the formula reads:

Kelly fraction = win rate − (loss rate ÷ payoff ratio)

The result is a share of the account. The math is clean. The inputs are not, as the sections below show. A larger edge gives a larger share. A higher payoff ratio lets a lower win rate still produce a positive result. If the edge is gone, the result drops to nothing, and if the edge is negative, the result turns negative too.

Note what the formula optimizes. It does not aim for a smooth equity curve, and it does not aim to keep any single drawdown small.

It aims for the highest growth rate of the account over a very long run of identical bets. Kelly maximizes growth, not comfort. The path it takes to get there can be brutal, and the formula has no opinion about that.

What it is not

Kelly is not the same thing as risking a fixed percentage per trade. Fixed-percentage sizing starts from a risk budget you choose and turns it into a size through your stop distance. That mechanic is explained in how to turn a risk amount and a stop distance into a size.

Kelly instead claims to tell you what the risk budget itself should be. That is a much bigger claim. It is only as good as the inputs behind it.

What the Formula Assumes

The Kelly result is exact only inside a model. Strip the model down, and it rests on a short list of conditions:

  • You know your true win rate and payoff ratio. Not an estimate. The real values.
  • Those values never change. Every trade comes from the same, stable distribution.
  • Trades are independent. A loss tells you nothing about the next trade.
  • Every outcome is a clean win or a clean loss of a fixed size. No partial exits, no gaps, no slippage.
  • You can re-size continuously. After every trade, the position is set as a share of whatever is left.
  • There is no floor. The account can fall as far as the sequence takes it, and play simply continues.

None of these holds on a real trading account. Not cleanly. Your win rate comes from a limited history.

Market conditions shift, so the edge you measured in the past may not be the edge you have now. Losses can come in clusters when a market changes character. Outcomes vary in size, costs come out of every fill, and fills are not always where you planned.

The last assumption matters most for a funded trader. In the formula's world, an account can keep falling forever and never quite empty, because every bet is a share of what is left. A real account has a line well before empty where trading stops. That is the catch.

Why Small Estimation Errors Make the Size Too Big

Here is the core problem. You never feed the formula your true win rate and payoff ratio. You feed it estimates, built from a sample of past trades.

Over a short run, chance alone can make a setup look better than it is. If you want to know why a handful of trades says little, the FAQ of the risk-reward guide covers it.

Estimation errors would be harmless if they cancelled out. They do not. The formula reacts strongly to the edge, and the damage is lopsided.

Start with the sensitivity. A modest overestimate of your win rate can raise the calculated share by more than the overestimate itself, because the edge is a thin margin between larger terms. And a setup with no real edge at all can still produce a positive Kelly result from an optimistic sample.

Then the asymmetry. Sizing below the true Kelly value costs you some growth, gently. Sizing above it adds swings much faster than it adds growth.

Past a certain size, larger bets do not just add risk; expected growth starts to fall, and further out it turns negative even with a genuine edge. Oversizing costs more than undersizing. So an estimate that is wrong in the optimistic direction hurts you more than one that is wrong in the cautious direction.

The practical reading is simple. Treat any Kelly result as an upper bound computed from inputs you cannot fully trust. The less data stands behind your win rate, the further below that bound the honest answer sits.

Risk of Ruin: The Probability of Hitting Your Limit

Kelly asks how to grow fastest. Risk of ruin asks a different question: what is the probability that the account reaches a set limit before it gets the chance to grow?

The limit is whatever line you cannot cross. On a personal account, it might be the drawdown at which you stop trading a strategy. On an evaluation or funded account, it is the loss limit written into the rules.

How such limits are measured, from a fixed reference point or a trailing one, is explained in how a drawdown limit is measured.

Risk of ruin moves with the same inputs as Kelly, but it answers to the limit. The mechanics run in clear directions:

  • A larger share per trade raises it. Fewer losses in a row are needed to reach the limit.
  • A larger edge lowers it. Winners arrive often enough to pull the account away from the line.
  • A tighter limit raises it. The same streak that a loose limit absorbs can end an account with a tight one.
  • Uncertain inputs raise it. If the edge is smaller than you think, every estimate built on it is too optimistic.

Risk of ruin only approaches nothing when the share per trade is small relative to the room above the limit. Your limit sets your real budget. That is the part Kelly leaves out.

This is also where risk of ruin differs from counting losing streaks. The risk-reward guide shows how to divide a limit by your risk per trade to see how many full losses in a row fit.

Risk of ruin is the probability version of the same question. It asks how likely such a streak is, given your edge and your size.

Why Kelly Ignores Your Loss Limit

The Kelly criterion assumes a world without a floor. It maximizes long-run growth on the assumption that you stay in the game no matter how deep the drawdown runs. Growth, not survival.

A trader on an account with a hard loss limit does not get that long run. The account is not slowed down at the line. It ends there.

That changes what a sensible size looks like. A share that maximizes growth across a very long run of trades can still produce a drawdown deep enough to cross a loss limit early. Under Kelly, a deep drawdown is a temporary dip on the way to growth. Under a hard limit, it is the end of the account.

On CarrotFunding, an account is breached when equity touches or falls below either of its equity limits, and those limits apply to realized and unrealized losses. The CarrotFunding Rulebook sets out how they work. Read them before you take any sizing formula at face value.

Fractional Kelly as a Concept

Fractional Kelly is one answer to these problems. The idea is to take only a fraction of the size the formula calculates.

A smaller share dampens the swings of the account and lowers the risk of reaching a limit. It also costs growth, because you are deliberately sizing below the point the model calls ideal.

Under uncertain inputs, that trade-off looks better than it does on paper. If your true edge is smaller than your estimate, a fraction of the calculated size may land closer to the true Kelly value than the full result would. Undersizing is the cheaper mistake, as the section on estimation errors showed.

Which fraction makes sense is not a question the formula answers. It depends on how much data stands behind your inputs and where your limit sits. This article does not suggest a value. What it does suggest is the order of the decision. The order matters. The limit and the risk you accept come first, and any sizing formula comes after.

A practical way to build that order is laid out in a framework for risk per trade, per day and across positions.

If the terms here are new, the learning paths and glossary on our learn page cover the basics in a fixed order.

Conclusion

The Kelly criterion is a precise answer to a question you are not quite asking. It sizes for maximum growth with perfect inputs and no floor, while you trade with estimated inputs and a hard line. So the decision in front of you is not whether to follow the formula. It is how much you actually know about your edge, and how much room your limit leaves. Start from the limit, judge how solid your win rate and payoff ratio really are, and read any Kelly result as a ceiling you stay well below. A precise number built on a guess is still a guess.

FAQ

Is the Kelly criterion a position-size recommendation?

No. It is a model output that is only correct when its assumptions hold, and in trading they do not. It can tell you that a size is too large for the edge you think you have. It cannot tell you what to trade, and it does not account for a loss limit, costs or changing market conditions.

What does a negative Kelly result mean?

It means the inputs show no edge, or a losing one. A result that drops to nothing says the expected gain and loss balance out, so any size only adds risk. A negative result says the setup loses money on average, and the formula would size the opposite side. In practice, it is a signal to stop and re-examine the setup, not to reverse it.

How is Kelly different from risking a fixed percentage per trade?

Fixed-percentage sizing starts from a risk budget you choose and converts it into a size using your stop distance. Kelly tries to derive the budget itself from your win rate and payoff ratio. Fixed-percentage sizing makes no claim about your edge, while Kelly depends on it entirely, which is exactly why its errors are larger.

Does risk of ruin ever reach zero?

Not on a real account with a limit. As long as losing trades are possible, there is some chance of a streak long enough to reach the line. A smaller share per trade and a larger edge push that chance down, but never quite to nothing. Uncertainty about your true edge means the real risk can be higher than a calculation built on your estimates shows.

Does the Kelly criterion apply to crypto perpetuals?

The arithmetic is the same on any market. On perpetual futures, fees and slippage come out of every trade, and funding payments add to or take from a position while it is held, so your real payoff ratio, and with it the Kelly result, can differ from the one you measured. Leverage changes how much margin a position ties up, not the share of the account you have at risk.

Contents
Topics

92 articles in total

Ready to trade with our capital?

Pass the evaluation and trade funded on a prop firm built by traders, for traders.

Get Funded

More Insights

Education

What Is MEV? Who Sees Your Transaction Before It Lands

MEV is the value someone can earn by choosing the order of transactions before they land in a block. Here is where your transaction waits, how front-running, sandwiches and back-running work, what changes between pools and order books, and why a public order is not the same as a fair one.

9 min read