Kelly Criterion: How to Size Bets Using Your Edge

What Is the Kelly Criterion?

The Kelly criterion is a staking formula that tells you what percentage of your bankroll to risk on a bet, based on two inputs: your estimated probability of winning, and the odds being offered. If you think a bet has a genuine edge — meaning the true win probability is higher than the odds imply — Kelly converts that edge into a specific stake size. For example, with a 55% win chance at evens (2.00) and a $1,000 bankroll, full Kelly says to stake $100.00 (10.00% of the bankroll). For a given set of odds, a larger edge generally means a larger stake; across different odds, the stake also depends on the odds/payout.

Why Bettors Use It

Betting the same flat amount every time ignores how confident you actually are in each bet. Betting too little on a strong edge wastes the opportunity; betting too much on any edge risks large bankroll swings. The Kelly criterion solves this by scaling your stake to the size of your edge, so stronger opportunities get proportionally larger stakes and weak or negative edges get little or nothing.

This matters because odds — whether shown as American odds, decimals, or fractions — already bake in a bookmaker margin, known as the vig. Beating that margin consistently is hard, and Kelly only makes sense once you have a genuine, well-reasoned probability estimate that clears it.

The Formula, in Plain English

Kelly’s core idea is:

Stake % = Edge ÷ Odds (in decimal terms, minus 1)

Where “edge” is the expected profit per unit staked: (estimated win probability × decimal odds) − 1. You don’t need to memorize the algebra — a Kelly criterion calculator does the math for you. What matters is understanding the two inputs that drive the output: This applies to a simple win-or-lose bet; markets that can push need an adjusted formula.

1. Your win probability estimate — a genuine, independently formed view of how likely the outcome is.
2. The odds on offer — which set the breakeven probability you need to beat.

Worked Example

Let’s say you believe a team has a 55% chance to win a game priced at evens (2.00), and you’re working with a $1,000 bankroll.

• Edge: 10.00%
• Full Kelly stake: 10.00% of bankroll
• Dollar stake: $100.00

Now compare that to a bet at American odds of +150, where you estimate a 45% win chance, also on a $1,000 bankroll:

• Edge: 12.50%
• Full Kelly stake: 8.33%
• Dollar stake: $83.33

Even though the edge is larger in percentage terms (12.50% vs. 10.00%), the recommended dollar stake is smaller ($83.33 vs. $100.00), because the odds and payout structure are different. This is why Kelly can’t be eyeballed reliably — the interaction between probability and odds isn’t always intuitive.

Here’s a third case: a bet at −110, where you estimate a 55% win chance, on a $1,000 bankroll.

• Edge: 5.00%
• Full Kelly stake: 5.50%
• Dollar stake: $55.00

And finally, a bet at −110 with only a 50% win chance — a coin-flip estimate against odds that already have a built-in margin:

• Edge: -4.55%
• Full Kelly stake: -5.00%
• Stake: 0.00%

A negative Kelly output means the math is telling you there’s no edge — the correct stake is nothing.

Full Kelly vs. Half Kelly

Full Kelly maximizes long-run theoretical growth, but it’s volatile. Because it stakes aggressively on every perceived edge, a string of losses — or a probability estimate that turns out to be wrong — can produce sharp bankroll swings. A common alternative is half Kelly: simply betting half of what full Kelly recommends.

Using the first example above (55% win chance, evens, $1,000 bankroll):

Approach	Stake %	Dollar Stake
Full Kelly	10.00%	$100.00
Half Kelly	5.00%	$50.00

Half Kelly roughly trades some long-run growth for meaningfully less volatility — a trade-off that can make sense, especially when probability estimates are uncertain (which they always are).

Common Mistakes

• Treating Kelly as precise. The formula is only as good as your probability estimate. A guess dressed up as a percentage is still a guess, and small errors in that estimate can swing the recommended stake significantly.
• Betting full Kelly without understanding the volatility. Full Kelly stakes can produce large bankroll swings even when the underlying edge is real.
• Ignoring negative results. If your inputs produce a negative edge — like the −110, 50% example above — the correct action is not to bet, not to round up to a small stake anyway.
• Applying Kelly across correlated bets, such as multiple legs of a parlay or accumulator, without adjusting for the fact that the outcomes aren’t independent.
• Forgetting the odds format matters. Whether odds are shown as American, decimal, or fractional, they need to be converted consistently before the formula is applied — see our guide on how betting odds work.

Is Kelly Right for You?

The Kelly criterion is a mathematical framework, not a guarantee. It assumes you can estimate probabilities accurately, which is genuinely difficult, and it assumes you’re comfortable with the swings that come with staking proportionally to your bankroll. Fractional Kelly (like half Kelly) is a common way specifically to soften those swings while still scaling stakes to perceived edge.

Nothing in this guide is financial advice, and a Kelly calculation is only ever as reliable as the probability estimate you feed into it. Use it as one tool among many for thinking about stake sizing, not as a guarantee of profit.

If you want to run your own numbers, our Kelly criterion calculator lets you plug in your estimated probability and the odds to see the suggested stake instantly.

Betting always carries risk, and no staking formula changes that. Please gamble responsibly and only bet what you can afford to lose (18+/21+ where applicable).