Betting maths · Free tool · No sign-up

Kelly Criterion Calculator

The Kelly criterion is a formula that turns a price and your own win-chance estimate into the share of a bankroll the model would stake. Half and quarter Kelly shrink that share to cut swings.

Kelly % = (p × odds − 1) ÷ (odds − 1)
Checked Nov 15, 2026 by LegalDeps Editorial Team
Try an example
Updates as you type · No sign-up

How this result is worked out

    What this tool does

    The Kelly criterion is a mathematical stake model based on estimated probability and a betting price. It aims to relate the size of a stake to an estimated advantage. Its answer depends heavily on the probability you supply. An inaccurate estimate can create a large suggested fraction where no real advantage exists. This calculator explains the model; it does not recommend a personal stake.

    How to use it

    Enter decimal odds, your estimated chance as a percentage, and a bankroll already set aside. Choose full, half or quarter Kelly. The tool shows the mathematical stake and the estimated edge. Full Kelly uses the entire positive model fraction. Half and quarter Kelly multiply that fraction by one half or one quarter. When the estimated edge is zero or negative, the displayed stake is zero.

    The formula and assumptions

    Let p be your estimated win probability and d the decimal price. The full Kelly fraction is (p multiplied by d minus one) divided by (d minus one). A negative fraction is set to zero here. Multiply the positive fraction by the chosen Kelly share and the bankroll. Edge is p multiplied by d minus one, shown as a percentage of the stake.

    Worked example one

    At decimal 2.00 and an estimated 60 percent chance, the edge is 20 percent. Full Kelly is also 20 percent of bankroll. For a bankroll of 1,000, that means 200 under this model. Half Kelly gives 100 and quarter Kelly gives 50. Those smaller figures still depend on the same uncertain probability estimate.

    Worked example two

    At decimal 3.00 and an estimated 30 percent chance, p multiplied by d is 0.90. The estimated edge is negative ten percent. The raw Kelly fraction is negative, so this tool shows zero. A larger potential payout does not create an advantage when the chance of winning is too low for that price.

    Common mistakes

    Do not substitute confidence, recent wins or the bookmaker implied chance for a well-founded probability estimate. A probability of 60 percent is entered as 60, not 0.60. Do not use your entire savings as a bankroll. The basic model assumes repeated opportunities with known probabilities. It does not account for correlated open bets, limits, taxes, commission or a changing price.

    Reading the result

    Even a mathematically positive Kelly estimate can produce sharp losses. Fractional Kelly reduces the modeled exposure but does not remove estimation risk. Test how the answer changes if your estimated chance is several points lower. If a tiny change eliminates the edge, the result is sensitive. A fixed affordable loss limit and taking a break matter more than an exact-looking model output.

    Read our betting guides for the surrounding rules. Use the glossary for unfamiliar words. You can also browse betting sites by country or casino lists to check where the local rules apply. These tools explain calculations and recorded facts. They do not make gambling income predictable.

    Frequently asked questions

    What is the difference between edge and Kelly fraction?

    Edge is the model expected profit relative to a stake. Kelly also divides by the net odds, producing a bankroll fraction. They are not generally equal.

    Why can the result be zero?

    A non-positive estimated edge produces no positive Kelly stake. Increasing the bankroll does not change the sign of that edge.

    Does half Kelly halve the probability?

    No. It halves the full Kelly stake fraction. Your entered probability and the estimated edge remain unchanged.

    Can the calculator estimate my chance?

    No. You must provide the probability. The page cannot observe teams, injuries, market information or the reliability of your model.

    Is full Kelly a safe limit?

    No. It is a model output under strong assumptions. Probability error, correlated bets and real-world restrictions can make it inappropriate.

    Should I use the result as betting advice?

    No. Use it to understand the formula and its sensitivity. It is not a personal recommendation or a promise of growth.

    What to do next

    Margin CalculatorCheck how much margin sits in the price first Hedge CalculatorBalance an open bet if the price moves