Kelly Criterion: Optimal Bet Sizing Explained

The Kelly criterion is a formula that calculates the fraction of capital to risk on a trade to maximize long-term account growth, given your win probability and your win/loss ratio. It is the mathematically optimal bet size for a known edge, but because traders rarely know their edge precisely, most use a fraction of what Kelly suggests.

What is the Kelly criterion?

The Kelly criterion, developed by John Kelly in 1956, answers a precise question: given a known edge, what fraction of your capital should you risk per bet to grow wealth fastest over the long run? Bet too little and you leave growth on the table; bet too much and the volatility, and the risk of ruin, eventually overwhelm the edge. Kelly finds the sweet spot between those failures.

For traders, Kelly reframes position sizing as an optimization rather than a fixed rule. Instead of always risking 1 percent, it suggests sizing up when your edge is strong and down when it is weak. That sounds ideal, but it hinges entirely on knowing your true edge, which is where the theory meets the messy reality of markets and why caution is built into how practitioners actually use it. The foundation it sits on is covered in position sizing strategies.

What is the Kelly formula?

The Kelly formula for trading uses two inputs: your win probability and your payoff ratio. It returns the fraction of capital to risk for maximum long-run growth.

Kelly Fraction = W - (1 - W) / R

Here W is the probability of a winning trade (as a decimal) and R is the ratio of your average win to your average loss. If W is 0.5 and R is 2 (winners twice the size of losers), Kelly = 0.5 minus (0.5 / 2) = 0.5 minus 0.25 = 0.25, suggesting you risk 25 percent of capital per trade. That number alone should signal why full Kelly is rarely used directly: 25 percent per trade is enormous risk that would produce violent swings. The formula's output is a ceiling to scale down from, not a target to take literally.

A worked example

Suppose a strategy wins 55 percent of the time, and its average winner is 1.5 times its average loser. Plugging into Kelly tells you the growth-optimal fraction, which you then deliberately reduce.

W = 0.55,  R = 1.5
Kelly = 0.55 - (0.45 / 1.5) = 0.55 - 0.30 = 0.25

Full Kelly suggests risking 25 percent of capital per trade, an amount almost no trader could stomach, since a few losses would gut the account. The practical move is fractional Kelly: half-Kelly risks 12.5 percent, quarter-Kelly risks 6.25 percent, and many traders go lower still. The table below shows how fractional Kelly trades a little growth for a lot less volatility.

Why use fractional Kelly?

Full Kelly is optimal only if your edge estimate is exactly right, and it is almost never exactly right. Overestimate your win rate or payoff even slightly and Kelly tells you to bet too much, which can turn a real edge into a blowup. Fractional Kelly, typically half or quarter, gives up a modest amount of theoretical growth in exchange for a large reduction in drawdown and a big margin for error in your inputs.

The reason this trade is worth it lies in the asymmetry. Studies of Kelly betting show that growth falls off slowly as you reduce below full Kelly, but drawdowns shrink quickly. Half-Kelly keeps roughly three quarters of the growth with far gentler swings, which most traders find far easier to follow without panic. Given that your edge estimate carries real uncertainty, betting below the theoretical optimum is not timidity, it is acknowledging that you do not know your true numbers.

What are the limits of the Kelly criterion?

Kelly's biggest limit is its assumption that you know your true probabilities, which traders almost never do. Edges drift, markets change regime, and a backtested win rate may not hold live. Feeding optimistic inputs into Kelly produces dangerously large sizes, the opposite of safety. This sensitivity is why beginners are usually better served by a simple fixed cap like the risk per trade rule.

Kelly also ignores correlation: it sizes each bet in isolation, but if you hold several correlated positions, your real risk is far higher than any single Kelly fraction implies. And it assumes you can emotionally tolerate full-Kelly volatility, which few can. For these reasons, treat Kelly as a conceptual ceiling and a way to think about scaling size with edge, not as a literal sizing engine. Combine its insight with a hard risk cap and verify any edge with honest expectancy in trading math.

How does Kelly compare to fixed percent risk?

Kelly and the fixed percent risk rule both produce a fraction of capital to risk, but they arrive there differently and suit different traders. Fixed percent risk picks a constant fraction (say 1 percent) by judgment and applies it to every trade regardless of edge. Kelly computes the fraction from your measured win rate and payoff, so it varies with the quality of each setup. One is a fixed safety rule; the other is an edge-optimizing target.

The practical difference shows up in their failure modes. Fixed percent risk fails only by being suboptimal: at 1 percent you may grow slower than you could, but you almost never blow up. Kelly fails by being fragile: if your edge estimate is wrong, it can size you into ruin. For a beginner without a reliable track record, the fixed rule's robustness is worth far more than Kelly's theoretical efficiency, which is why our risk per trade rule guide recommends it as the default.

A sensible progression is to start with fixed percent risk, build a real sample of trades, and only then consider letting a conservative fractional Kelly nudge your size up on your strongest, best-confirmed setups. Even then, keep a hard cap so Kelly's output can never exceed a level you are comfortable losing. The two methods are not rivals so much as stages: the fixed rule keeps you alive long enough to earn the data Kelly needs.

Putting Kelly in context

The Kelly criterion is most valuable as a way of thinking: bet more when your edge is real and strong, less when it is thin, and never so much that variance can ruin you. Used literally with uncertain inputs, full Kelly is reckless; used as a scaled-down guide alongside a firm risk cap, its logic improves how you allocate risk across setups of different quality.

For most traders, the sensible path is a small fixed percent per trade until you can estimate your edge from a real track record, then optionally lean toward a conservative fractional Kelly for higher-conviction setups. The arithmetic is easy to run, but the discipline, sizing below the optimum and respecting your own uncertainty, is what keeps it safe. Run the practical numbers with the position size calculator.