Kelly Criterion for Betting: How Much Should You Stake?
Updated 2026-10-026 min read
The Kelly criterion is a formula that gives the fraction of your bankroll to stake on a bet with a known edge, in order to maximise long-term growth. It is powerful and dangerous at the same time: it assumes your probability is exactly right, and real estimates never are. That is why most practitioners use a fraction of it.
The formula
f = (b × p − q) ÷ b, where b = decimal odds − 1, p = probability of winning, q = 1 − p. If f is zero or negative, do not bet.
Worked example
Odds 2.20, so b = 1.20. Estimated probability p = 50%, q = 50%. f = (1.20 × 0.50 − 0.50) ÷ 1.20 = 0.10 ÷ 1.20 = 8.3% of the bankroll.
On a bankroll of 1,000 that means a stake of about 83. Half Kelly would be 4.2%, about 42.
Why fractional Kelly
If the real probability is 45% instead of 50%, the same bet at 2.20 has an expected value of −1%, and full Kelly would have you staking on a losing proposition. Over-estimating your edge makes full Kelly punish you hard, while under-estimating costs only a little growth.
- Half or quarter Kelly gives up some growth in exchange for far smaller drawdowns.
- A hard cap on the stake per bet (for example 2-3% of the bankroll) protects against estimation error.
- Kelly applies to one bet at a time. Simultaneous bets on correlated events need a more careful treatment.
No staking plan can turn a negative-expectation bet into a profitable one. Never stake money you cannot afford to lose.
Frequently asked questions
What is the Kelly criterion in betting?
A formula that gives the fraction of your bankroll to stake to maximise long-term growth, based on your estimated probability and the odds.
What is the Kelly formula for sports betting?
f = (b × p − q) ÷ b, where b is decimal odds minus 1, p the win probability and q = 1 − p.
Is full Kelly too risky?
Usually yes. Since the probability is an estimate, most bettors use half or quarter Kelly to reduce volatility and the cost of estimation errors.
More guides
AlphaBetAI scans the markets with an ensemble of models and publishes a verifiable track record of every graded result.
Explore AlphaBetAI