The Kelly Criterion is a mathematical formula that tells you how much of your bankroll to bet or invest on a single opportunity, based on your edge. Developed by Bell Labs scientist John Kelly Jr. in 1956, it was originally designed for information theory problems but quickly found a home in gambling and investing. The core idea: bet enough to maximize long-term growth, but not so much that a losing streak wipes you out.
The Formula and What Each Variable Means
The standard Kelly Criterion formula looks like this:
Kelly % = W − [(1 − W) / R]
The result, Kelly %, is the percentage of your total capital you should put into a single bet or trade. W is your win probability, meaning the likelihood that any given bet or trade will be profitable. R is your win/loss ratio, calculated by dividing your average winning amount by your average losing amount.
Say you have a trading strategy that wins 60% of the time, and your average winner is $300 while your average loser is $200. Your W is 0.60 and your R is 1.5 ($300 / $200). Plug those in:
Kelly % = 0.60 − [(1 − 0.60) / 1.5] = 0.60 − 0.267 = 0.333
The formula says to allocate about 33% of your capital to each trade. That feels aggressive, and it often is. We’ll get to why most practitioners scale it down.
Why It Maximizes Long-Term Growth
The Kelly Criterion isn’t trying to maximize your profit on any single bet. It maximizes the expected growth rate of your bankroll over many repeated bets. This distinction matters. If you bet too little, you leave money on the table. If you bet too much, a string of losses can devastate your account, and the math of recovery works against you (a 50% loss requires a 100% gain just to break even).
Kelly finds the sweet spot between these two extremes. Over a long enough series of bets where you have a genuine edge, no other fixed-fraction strategy will grow your capital faster. This property is sometimes called “maximizing the geometric mean” of returns. In plain terms, it compounds your money as efficiently as possible without taking on so much risk that bad runs destroy your progress.
One important detail: the formula assumes you’ll reinvest your winnings and keep betting a percentage of your updated bankroll each time, not a fixed dollar amount. As your bankroll grows, your bet sizes grow proportionally. As it shrinks, your bets shrink too, which is a built-in form of risk management.
Using Kelly for Investing
The formula was built for situations with clear, binary outcomes: you win or you lose a known amount. Gambling fits that mold perfectly. Financial markets are messier. Stock prices can move by any amount in either direction, returns aren’t binary, and the probabilities are much harder to pin down.
To apply Kelly to investing, you estimate W and R from your trading system’s historical performance. W becomes the percentage of trades that were profitable, and R becomes the ratio of average gains to average losses across your track record. The formula itself doesn’t change, but the inputs shift from theoretical probabilities to historical statistics.
Some investors adapt the concept further for portfolio allocation. Instead of asking “how much should I bet?”, they ask “how much of my portfolio should go into this particular stock or asset?” The logic is the same: allocate more to opportunities where your edge is larger, less to opportunities where it’s smaller or more uncertain.
Why Most People Use Fractional Kelly
Full Kelly sizing is mathematically optimal for long-term growth, but it comes with stomach-churning volatility along the way. In the example above, putting 33% of your capital into a single trade means a few consecutive losses could cut your account in half. Most people, whether professional gamblers or portfolio managers, can’t tolerate that level of drawdown even if the math says they’ll recover eventually.
The practical solution is fractional Kelly. Half-Kelly means you take whatever the formula recommends and cut it in half. Quarter-Kelly cuts it to 25% of the full recommendation. In our example, half-Kelly would mean allocating about 16.5% per trade instead of 33%.
The tradeoff is straightforward. Half-Kelly reduces your maximum long-term growth rate, but it also dramatically reduces the size and frequency of painful drawdowns. For most people, the smoother ride is worth the slightly slower compounding. Many experienced practitioners treat full Kelly as a theoretical ceiling and operate somewhere between 25% and 50% of the full recommendation.
The Estimation Problem
The Kelly Criterion’s biggest practical weakness has nothing to do with the math itself. It’s that the formula is only as good as the numbers you feed it. Even small inaccuracies in your estimates of win probability or the win/loss ratio can lead to significant over-investing or under-investing.
This is known as estimation risk. If the real-world parameters differ from the ones you used to calculate your position size, the growth rate of your portfolio can become “strongly suboptimal,” as researchers have documented. In the worst case, overestimating your edge leads you to bet far too aggressively, which can actually produce negative expected growth, turning a winning strategy into a losing one through sheer over-sizing.
In gambling, the probabilities might be reasonably well-known (card counting in blackjack, for instance, provides fairly precise edge estimates). In investing, your estimates of future win rates and payoff ratios are inherently uncertain. Past performance may not reflect future conditions. This uncertainty is another reason fractional Kelly is popular: by betting less than the formula suggests, you build in a margin of safety against the possibility that your inputs are wrong.
When Kelly Works Best
The Kelly Criterion is most useful when three conditions are met. First, you have a quantifiable edge. If you can’t estimate your win probability and average payoffs with reasonable confidence, the formula gives you a precise answer to an imprecise question. Second, you’re making the same type of bet repeatedly over time. Kelly is a long-run strategy; on any single bet, it doesn’t protect you from losing. Third, you’re reinvesting gains and can tolerate some volatility, since the formula assumes your bet size adjusts with your bankroll.
Professional sports bettors, poker players, and systematic traders are the most common users. Some hedge funds and quantitative investors apply Kelly-based frameworks to portfolio construction, though usually with significant modifications and conservative adjustments. For a casual investor buying index funds for retirement, Kelly is interesting to understand but rarely something you’d apply directly, since the inputs are too uncertain and the framework assumes a level of precision that long-term, diversified investing doesn’t require.
Where the Kelly Criterion genuinely shines is as a mental model. Even if you never plug numbers into the formula, the core insight is valuable: the size of your bet should be proportional to the size of your edge. When your advantage is large and well-understood, bet more. When it’s small or uncertain, bet less. That principle applies whether you’re allocating a portfolio, making business investments, or deciding how much to wager on a poker hand.