Enter the decimal odds, your estimate of the win probability, and your current bankroll. The calculator returns the full Kelly stake as a percentage of that bankroll and in cash, along with the half and quarter Kelly alternatives. Kelly scales the stake to the size of your edge, so a thin edge produces a small bet.
How Kelly Sizes a Bet
Kelly answers a staking question rather than a which-bet question. You bring the price and the probability, and it returns the fraction of bankroll that grows the bank fastest across a long run of bets at that kind of edge.
f* = (b × p – q) ÷ b
b is the profit per unit staked, so decimal odds of 2.50 give b = 1.5. p is your win probability and q is whatever remains after it. The numerator is your edge in cash terms, and dividing by b shrinks the fraction as the price lengthens, so the same edge at a long price earns a smaller stake than at a short one.
Overbetting past f* slows long-run growth and deepens the drawdowns at the same time. Staking under it costs some growth and buys a quieter equity curve, a trade most bettors make on purpose.
A Worked Example
Decimal odds of 2.50, a win probability estimate of 45%, and a bankroll of $1,000.
- b, the profit per unit staked: 2.50 – 1 = 1.5.
- Full Kelly fraction: (1.5 × 0.45 – 0.55) ÷ 1.5 = 0.0833, or 8.33% of bankroll.
- Full Kelly in cash: $83.33.
- Half Kelly: 4.17% of bankroll = $41.67.
- Quarter Kelly: 2.08% of bankroll = $20.83.
Hold the price at 2.50 and drop the estimate to 38%. Breakeven at 2.50 is 40%, so 38% sits on the wrong side of the line and the formula returns -3.33%. A negative fraction is the no-bet result. No smaller stake rescues a price below fair value, and the correct size is zero.
Reading the Three Stake Lines
Each figure comes back as a percentage of the bankroll you entered, which is what keeps the plan self-correcting. The percentage holds steady while the cash stake falls after a losing run and climbs again as the bank recovers.
Full Kelly is the mathematical ceiling. Anything above 8.33% in this example is overbetting by definition, and the growth rate starts falling once you cross it. Half and quarter Kelly are the same recommendation scaled down to $41.67 and $20.83.
The 8.33% traces back to a 5-point edge: 45% against the 40% breakeven the price implies. Run the same inputs through the Expected Value Calculator and you get that edge priced in cash, while the Implied Probability Calculator gives you the 40% on its own.
Full Kelly Punishes an Optimistic Estimate
8.33% of a bankroll on one football match is a large bet by most standards, and it only earns that size if the 45% is right. Suppose the true chance is 40%. The bet now carries no edge at all and you have $83.33 of a $1,000 bank riding on it. Kelly multiplies your stake by your confidence, so an inflated probability produces an inflated bet.
That error does not announce itself. A 45% estimate on a 40% bet still wins nearly half the time, and the damage shows up as a slow bleed across dozens of bets rather than as one obvious mistake.
Half and quarter Kelly exist for that reason. Halving the fraction halves the loss on every loser, $41.67 rather than $83.33, and it buys room to be wrong about your probabilities without wrecking the bank. Most bettors who use Kelly run a fraction of it, and the ones who run full Kelly on estimates they built themselves tend to discover the difference between a model and an opinion the expensive way.
One habit worth keeping: recalculate against your current bankroll, not the figure you started the season with.
Where the Probability Comes From
Gecko Edge was built for the input rather than the arithmetic. The platform produces a win probability for every market through Poisson goal models with a Dixon-Coles correction, blended with market and league priors, across 130+ leagues. That number is the one Kelly needs, and it is the one most staking plans are missing.
Further Reading
- All betting calculators: the full library of 34 free tools
- Losing runs and variance: what to expect
- Stop guessing, start testing: backtesting your betting systems
What is the Kelly Criterion in betting?
Kelly is a staking formula that sets your bet as a fraction of bankroll based on the size of your edge. It uses f* = (b × p – q) ÷ b, where b is profit per unit staked at decimal odds, p is your win probability and q is 1 – p. At odds of 2.50 with a 45% estimate it returns 8.33% of bankroll.
Is half Kelly better than full Kelly?
Half Kelly stakes half the recommended fraction: $41.67 rather than $83.33 on a $1,000 bank in the example above. It gives up some long-run growth and cuts the depth of the drawdowns, and it leaves room for your win probability to be a little optimistic without doing serious damage. Quarter Kelly halves the exposure again.
What does a negative Kelly stake mean?
A negative fraction means the price is below fair value for your estimate, so the calculator returns no bet. At odds of 2.50 with a 38% estimate the formula gives -3.33%, because 2.50 needs a 40% win rate to break even. Kelly has no mechanism for backing a losing price at a reduced stake.
What win probability should I enter into a Kelly calculator?
Your own honest estimate, settled before you look at the price so the odds do not anchor it. Most bettors get this from a statistical model, from sharp market prices stripped of margin, or from a modelling platform. Gecko Edge generates one for every market from a Poisson and Dixon-Coles pipeline before any analysis appears.