The Kelly formula
The fraction to stake is (bp minus q) divided by b, where b is the decimal odds minus one, p is your probability of winning and q is one minus p. At odds of 3.0 with a 40% chance: b is 2, so (0.8 minus 0.6) divided by 2 equals 0.10, a tenth of the bankroll. If the result is zero or negative there is no edge and the Kelly stake is nothing.
Full Kelly is aggressive, and most people who try it once do not try it twice. It assumes your probability is spot on. Overestimate it a little and you overstake a lot, and the swings get ugly. Most practitioners use half or quarter Kelly, which gives up a little growth for far less variance, and the calculator offers those fractions directly.
Limitations
Kelly sizes one bet at a time. With several bets open at once the fractions should be reduced further, and the formula takes your probability estimate on trust. That estimate is where things go wrong. Use the output as an upper limit.
Worked example
A £1,000 bankroll, odds of 2.20 and your estimate of 50%. Full Kelly is (1.2 × 0.5 minus 0.5) divided by 1.2, which is 8.3%, a stake of £83.33. Half Kelly is £41.67. At the same odds with a 45% estimate the edge is negative and the recommended stake is zero.
Frequently asked questions
Why does Kelly recommend such large stakes?
Because it optimises growth rate, not comfort, and assumes your edge is real and known precisely. Neither is true in real betting, so most people stake a fraction of it.
Can I use Kelly for accumulators?
Mathematically yes, with the acca's combined odds and your probability that every leg wins. The trouble is that the probability estimate for a multi-leg bet is so uncertain that the output is rarely useful.
