What is the Kelly criterion?
Kelly is a formula for how much to stake, given a price and your own estimate of the probability. It answers "how much", never "whether" — and it assumes the estimate you feed it is correct, which is where almost all of the trouble comes from.
What is the Kelly criterion?
Published by John Kelly at Bell Labs in 1956, the criterion identifies the stake size that maximises the long-run growth rate of a bankroll when you have a repeatable edge. Stake too little and the bankroll grows more slowly than it could; stake too much and the swings eventually take it apart, even when every individual bet was sound.
The essential thing to understand before using it is what it does not do. Kelly does not find value, does not check your reasoning, and has no opinion about whether the number you gave it is realistic. It converts a claimed edge into a stake. If the edge is imaginary, the formula will size it just as confidently.
What is the Kelly formula for decimal odds?
p = your estimated probability · d = decimal odds
Suppose you estimate an outcome at 55% and the price on offer is 2.00. Then (0.55 × 2.00 − 1) ÷ (2.00 − 1) = 0.10 ÷ 1 = 0.10 — Kelly recommends staking 10% of the bankroll. On a $1,000 bankroll that is $100.
If the formula returns zero or a negative number, it is telling you something useful: at your own estimate, this price carries no edge, so there is no stake worth making. A negative Kelly is not an instruction to bet the other side — it is an instruction not to bet.
Our free Kelly criterion calculator does this arithmetic, including fractional Kelly.
Why do bettors use fractional Kelly?
Because the formula is far more sensitive to your probability estimate than it first appears. Return to the example: at a true 55% the correct stake is 10% of bankroll. If you were actually a little optimistic and the outcome is really 52%, the correct stake was 4% — you have staked two and a half times too much. At a true 50% the correct stake is nothing at all.
Since nobody's estimates are exact, staking the full number means routinely overbetting by an unknown margin. Half-Kelly or quarter-Kelly surrenders a modest amount of theoretical growth in exchange for a large reduction in volatility and in the chance of destroying the bankroll during a normal losing run. Most disciplined bettors who use Kelly at all use a fraction of it.
Does the Kelly criterion guarantee anything?
No. It optimises growth under assumptions that rarely hold in practice: that your probability is accurate, that the same edge is available repeatedly, and that you can stake whatever size the formula returns. Real betting breaks all three — estimates are uncertain, edges come and go, and bookmakers limit accounts.
Kelly is a discipline for sizing, not a method for winning. Applied to bets with no edge it does not protect you; it loses money at the mathematically optimal rate. The prior question — whether the price is actually worth taking — belongs to expected value and to removing the vig first. Nothing here is betting advice.
Where the estimate comes from
Everything above depends on p — the probability you supply. That is the hard part, and no staking formula can supply it for you. HaterPicks publishes a model projection beside each market across 79 Australian bookmakers, and grades every projection against the closing line so the estimate can be checked rather than trusted.
It is information and tooling, not betting advice. Read how HaterPicks works or try the Kelly calculator.