Compare the best odds
See the top list or jump straight to an operator that suits your playing style.
Every serious staking discussion eventually arrives at the same question: given an edge, how much of the bankroll should actually go on the bet? Betting too little leaves growth on the table over hundreds of wagers. Betting too much survives a handful of wins and then erases the bankroll on the losing run that always eventually arrives. The Kelly criterion is the formula that answers the question directly, and it is worth understanding properly rather than just typing numbers into a box β especially for a bettor who moves between odds formats, currencies and licensing regimes, where it is easy to lose track of what the fraction actually represents.
What the Kelly criterion actually solves
The Kelly criterion answers one specific question: what fraction of a bankroll, staked on a bet with a known edge, maximises the long-run growth rate of that bankroll? It does not answer "will this bet win" β that is a separate estimation problem the bettor has to solve first. Kelly takes a win probability as an input and tells you what to do with it.
The formula, in the form every online calculator uses, is:
- b = decimal odds β 1 (the net odds β how much profit one unit of stake returns if the bet wins)
- p = your estimated win probability
- q = 1 β p
- f* = (bΒ·p β q) / b
f* is the fraction of the bankroll to stake. When f* comes out at zero or below, the formula is telling you something specific: at your estimated probability, this price does not clear break-even, and the correct stake is nothing β not a small stake, nothing.
A worked example
Say a bettor prices a football match at 55% to go their way and finds decimal odds of 2.10 available on it.
- b = 2.10 β 1 = 1.10
- p = 0.55, q = 0.45
- f* = (1.10 Γ 0.55 β 0.45) / 1.10 = (0.605 β 0.45) / 1.10 = 0.155 / 1.10 β 0.1409
That is roughly 14.1% of the bankroll on a single bet β full Kelly. On a β¬1,000 bankroll that is a β¬140.91 stake. Most bettors who actually see that number for the first time flinch, and the flinch is correct: full Kelly is mathematically optimal for growth but brutally volatile in practice, because the formula assumes the probability estimate is exactly right, and no bettor's estimate is exactly right every time.
Why almost nobody bets full Kelly
The standard adjustment is fractional Kelly β betting a fixed fraction of what the full formula recommends, most commonly half or a quarter.
- Half Kelly on the example above: 7.05% of bankroll, β¬70.45 on the same β¬1,000 bankroll.
- Quarter Kelly: 3.52%, β¬35.23.
Halving the stake does not halve the growth rate β it cuts variance far more than it cuts expected growth, because the growth-rate curve is flat near its peak and steep in the tails. In practice this is the single most useful fact about Kelly staking: a bettor who consistently uses half Kelly gives up only a small amount of theoretical long-run growth in exchange for a dramatically smoother bankroll curve, which matters because bankrolls that survive downswings are the ones still betting when the edge eventually pays off.
The part the formula cannot do for you
Kelly staking is only as good as the win probability fed into it, and that number is a genuine estimate, not a fact you can look up. The bookmaker's price already tells you their implied probability β 1 divided by the decimal odds β and that is the number your own estimate has to beat, not just approximate. In the example above, 2.10 implies 47.6%; the bettor's edge exists only because they believe the true probability is 55%, not 47.6%. Overconfidence in that estimate is the most common way Kelly staking goes wrong in the real world: the formula is exact, the input rarely is, and a bettor who is systematically five percentage points too optimistic will be recommended stakes that are too large across every bet they place.
Kelly across odds formats and currencies
The formula needs a decimal price. American money-line odds (+150, β200) and fractional odds (6/4, 4/6) express the same information differently, and converting them incorrectly by hand is a common source of staking errors for bettors who shop between books that quote in different formats. Run the price through an odds converter first and feed the resulting decimal figure into the Kelly formula β never estimate a decimal equivalent in your head under time pressure.
Currency works differently from odds format: the Kelly fraction is currency-agnostic. Fourteen percent of a bankroll is fourteen percent whether the bankroll is held in euros, forints or lei β the fraction does not change when you convert. What changes is the final stake amount, and that is where cross-border bettors lose value quietly: converting a stake through a poor exchange rate or a card-network markup after the Kelly fraction has already been calculated erodes the edge the whole exercise was built to capture. Calculate the fraction first, then convert the resulting amount using the best available rate β not the other way round.
Try the calculator
The Kelly calculator takes bankroll, decimal odds and your estimated win probability, and returns the raw fraction plus full, half and quarter Kelly stakes side by side, so the size of the trade-off is visible before you commit to one. It flags non-value bets explicitly rather than ever returning a negative stake β if your inputs describe a bet with no edge, the tool tells you that directly.
Three mistakes that undo the formula
- Full Kelly on a probability estimate you are not confident in. The formula has zero margin for estimation error built in; fractional Kelly is the margin.
- Feeding it American or fractional odds directly. The math silently breaks if b is calculated from the wrong number base β convert to decimal first.
- Applying single-bet Kelly to an accumulator. Combined bets have correlated and compounding risk the single-event formula was never built to size; treat each leg's edge separately before combining stakes.
Responsible staking discipline matters more than any formula. If betting stops being a controlled, bankroll-managed activity and starts feeling compulsive, the responsible gambling page lists independent support options, including Gambling Therapy, that exist specifically for this.
FAQ
Is full Kelly ever the right choice?
Only for a bettor with a large bankroll, a long time horizon, and genuine confidence that their probability estimates are unbiased. For almost everyone else, the variance of full Kelly is worse in practice than the growth it theoretically buys.
What if my estimated probability is lower than the bookmaker's implied probability?
Then f* comes out at zero or negative, and the correct stake is nothing. The formula is doing its job β it is telling you that, at your own estimate, this is not a value bet.
Does Kelly work for casino games or only sports betting?
The formula assumes you can estimate a genuine edge, which is possible in sports and poker but essentially never true in casino games against the house β nearly every casino game carries a fixed negative edge for the player by design, so Kelly staking does not apply there in any useful sense.
How often should I recalculate my stake?
Every bet, using the current bankroll β not the bankroll from a week ago. Kelly staking compounds, and a stake sized off a stale bankroll figure drifts away from the formula's intent bet by bet.
For readers comparing operators across licences, our Ireland and Hungary hubs verify which sportsbooks actually accept players from each market before listing them, which matters more than any staking plan if the operator will not pay out in the first place.
Editorial note
Written by the Grand Bonuses editorial team from operator terms, regulator registers and our own country-by-country checks. Operators do not edit their reviews. Gambling is paid entertainment; play with a limit. Read more about our editorial process and our guidelines for responsible gaming.
