The Kelly criterion answers a narrow question: given a known edge and known odds, what fraction of your capital should you risk on each bet to maximize long-run growth? Although the formula is simple and the math is sound, the answer it gives is usually far too aggressive for a real trading account.
Full Kelly maximizes expected logarithmic growth, but it does so at the cost of enormous drawdowns and a high chance of ruin along the way. Most traders and investors are better served by risking a fraction of what the formula suggests.
This article walks through the formula, explains why full Kelly breaks down in practice, and shows how fractional Kelly fits into a risk-first approach.
The Kelly formula in plain terms
The classic Kelly formula for a binary bet with two outcomes is:
f* = (bp – q) / b
Here f* is the fraction of your capital to risk, b is the net odds you receive (the profit per unit staked if you win), p is your probability of winning, and q is the probability of losing, which equals 1 – p. If you win $2 for every $1 risked, b is 2. If you win $1 for every $1 risked, b is 1.
A quick example: suppose you have a strategy that wins 55% of the time and pays 1:1. Then p is 0.55, q is 0.45, and b is 1, so the formula gives f* = (1 × 0.55 – 0.45) / 1 = 0.10. Kelly says risk 10% of capital per bet.
For a continuous or multi-outcome setting, the formula generalizes. When you have an expected return and a variance, the Kelly fraction is approximately the expected excess return divided by the variance.
That version is common in portfolio construction, where you estimate the mean return of an asset or strategy and its volatility.
The appeal is obvious because Kelly is derived from a growth objective, not from gut feel. It tells you that risking too little leaves growth on the table, while risking too much eventually destroys capital; somewhere in between sits a maximum.
Why full Kelly is too aggressive
The formula assumes you know p and b exactly. In trading, you do not. Your win rate and payoff ratio are estimated from a finite sample, and they drift as markets change. A small error in your estimate of the edge produces a large error in the Kelly fraction.
Overestimate your edge and Kelly tells you to bet more than the true optimum, which pushes you past the peak of the growth curve and into the zone where more risk means less growth and more ruin risk.
Even with perfect estimates, full Kelly is brutal to live with. The math of log-optimal growth implies drawdowns that most people cannot tolerate. A common result is that full Kelly produces significant drawdowns at some point, and the expected time to recover can be long.
The theoretical growth rate is only achieved if you can sit through those swings without changing your behavior.
That last point matters more than the math because full Kelly assumes you will keep betting the same fraction after a significant drawdown. Real traders cut size, second-guess the system, or quit, so the behavioral failure turns a theoretically optimal strategy into a practically poor one.
There is also the estimation problem of fat tails, since markets produce losses larger than your model expects. As a result, a strategy calibrated on normal returns can hand you a single loss that wipes out many wins.
Kelly does not protect you from model error; instead, it amplifies the consequences.
Fractional Kelly and how to size it
Fractional Kelly means betting a fixed fraction of the full Kelly amount. If full Kelly says 10%, half Kelly says 5%, quarter Kelly says 2.5%. The fraction is usually written as a multiplier, so half Kelly is 0.5 × f*.
The trade-off is straightforward. Reducing the fraction lowers expected growth, but it lowers volatility and drawdown by more. Half Kelly gives up roughly a quarter of the growth rate while cutting the variance substantially.
Quarter Kelly gives up more growth but produces a much smoother equity curve. The exact numbers depend on the strategy, but the direction is reliable: less than full Kelly buys a large reduction in pain for a modest reduction in return.
How do you choose the fraction? A practical approach is to start from the estimate you trust least. If your edge estimate is shaky, use a smaller fraction. Many risk frameworks use a fraction of Kelly or less for strategies with uncertain parameters.
If you have a long, stable track record and a clear edge, half Kelly is still a defensible ceiling for most people.
Size the fraction to your drawdown tolerance, not to your greed. Work out what a significant account drawdown would do to your ability to keep following the plan. If the answer is that you would stop, your fraction is too high regardless of what the formula says.
Where the formula breaks in practice
Kelly assumes independent bets with fixed odds, yet trading rarely works that way. Positions overlap, correlations spike in a crisis, and several trades can lose together for reasons that are not independent.
Consequently, if you size each position as if it were isolated, your total risk can be far above what Kelly intended.
Kelly also assumes you can repeat the bet many times. A trader with a small number of trades per year faces sequence risk that the long-run formula smooths away. A string of losses early can end the account before the long-run average has a chance to show up.
There is a subtler issue with how the formula treats time. The Kelly fraction depends on the horizon and on how often you rebalance. A position sized for a multi-month hold is not the same as one sized for a two-week swing.
Mixing the two without adjusting the inputs produces a number that looks precise but means little.
Finally, Kelly says nothing about the quality of the edge. A positive expected value from a strategy that blows up in tail events is not a safe edge. The formula will happily tell you to size up a strategy that has a small chance of a catastrophic loss.
Risk management has to sit above the sizing formula, not inside it.
What this means for traders
If you trade on a horizon of weeks to a few months, your edge estimates are noisy and your sample of trades is small. Treat full Kelly as an upper bound you never touch. Estimate your win rate and payoff ratio from your own records, then apply a fraction well below one.
Quarter Kelly may be a starting point for a strategy you are still validating.
Watch correlation across open positions, because two trades that each look fine on their own can combine into a position larger than your Kelly budget. Cap total exposure at the portfolio level, not just per trade.
Recalculate the fraction as your sample grows, because early estimates are unreliable. Size small and let the data accumulate, so if the edge disappears, the Kelly fraction goes to zero or negative, which means no trade. That is a feature, not a failure.
What this means for investors
If you invest on a horizon of six months or more, the Kelly logic shows up in asset allocation rather than in single trades. The mean-variance version, expected excess return divided by variance, gives a rough guide to how much of a risky asset a growth-maximizing portfolio would hold.
Most long-term investors should hold less than that. The inputs, expected returns and volatility, are estimates with wide error bars, and the cost of being wrong is a drawdown you may not tolerate.
A fraction of the Kelly allocation leaves room for estimation error and for the fact that your risk tolerance is not a constant.
Rebalancing matters here because Kelly assumes you resize after each outcome. In a portfolio, that means trimming winners and adding to losers on a schedule, which is uncomfortable and often tax-inefficient. Therefore, factor those costs into the fraction you choose.
Conclusion: use Kelly as a ceiling, not a target
The Kelly criterion is a useful tool for thinking about position size. It tells you that there is an optimum and that oversizing is as harmful as undersizing. It does not tell you to bet the full amount it calculates.
For most traders and investors, fractional Kelly is the practical answer. Use the formula to find the ceiling, then size well below it to account for estimation error, fat tails, correlation, and your own tolerance for drawdowns. The goal is not maximum theoretical growth.
It is staying in the game long enough for your edge to matter.
- Understanding Drawdown: How Traders Recover From Losses
- Trailing Stops Compared: Percentage, ATR and Chandelier Exits
- Why the Risk-Reward Ratio Is Essential for Profitable Trading
Frequently Asked Questions
What is the Kelly criterion formula for trading?
For a binary bet, the formula is f* = (bp – q) / b, where b is the net odds, p is the win probability, and q is the loss probability. The result is the fraction of capital that maximizes long-run growth. For portfolios, a common approximation is expected excess return divided by variance.
Why is full Kelly too aggressive for real accounts?
Full Kelly assumes you know your edge exactly and can tolerate large drawdowns without changing behavior. In practice your estimates are noisy, losses can cluster, and most people cannot sit through the swings full Kelly produces.
Sizing below full Kelly reduces drawdowns far more than it reduces growth.
What is fractional Kelly?
Fractional Kelly means risking a fixed fraction of the full Kelly amount, such as half or quarter Kelly. It lowers expected growth but cuts volatility and drawdown substantially, which makes it easier to follow consistently.
How do I choose a Kelly fraction?
Start from how much confidence you have in your edge estimate and how large a drawdown you can tolerate without abandoning the plan. Strategies with short or noisy track records call for smaller fractions, often quarter Kelly or less.
Does Kelly work when trades are correlated?
Not directly. The formula assumes independent bets with fixed odds. Correlated positions can combine into more risk than the per-trade calculation suggests, so you need a portfolio-level exposure cap on top of the Kelly fraction.
AI Notice: This article was created wholly or predominantly with the assistance of artificial intelligence and was published without human editorial review.
This article is for general information only and does not constitute investment advice. Always do your own research before making trading decisions.