Should I use full Kelly or fractional Kelly?

Almost always fractional. Full Kelly assumes your win-probability estimates are exact, which they never are. Half or quarter Kelly cuts your bankroll volatility dramatically while keeping most of the long-run growth, and it forgives the inevitable errors in your probability estimates.

What happens if I bet more than Kelly suggests?

Over-betting relative to Kelly increases growth up to a point, then sharply increases your risk of ruin. Bet at roughly double the Kelly fraction and your expected long-run growth actually falls to zero despite every bet being +EV — the variance eats the edge.

Does Kelly work if I bet several games at once?

Standard Kelly assumes one bet resolves before the next. With simultaneous or correlated bets, staking each at full Kelly over-exposes your bankroll. Scale your fractions down, or size the group as a portfolio, when bets overlap in time or outcome.

How accurate do my probabilities need to be?

Accurate enough that your edge is real, not imagined. Kelly amplifies both good and bad estimates, so a bettor who overrates their edge will systematically overstake. Sanity-check your estimates against closing line value before trusting them for stake sizing.

Advanced7 min read

Kelly Criterion for Sports Betting: The Complete Guide

A formula named after a Bell Labs physicist sounds intimidating — it stopped me from touching it for a long time. Then I learned it in an afternoon, and it changed how I stake every bet. The Kelly criterion tells you what fraction of your bankroll to put on a bet, based on your edge and the odds. Used correctly it maximises the long-term growth of your bankroll while mathematically protecting you from going broke. Used carelessly — at full size, with sloppy probability estimates — it can be brutally volatile. This guide covers both sides.

The formula

For a simple win/lose bet at decimal odds, Kelly says stake this fraction of your bankroll:

f = (b × p − q) ÷ b

Where b is the decimal odds minus 1 (your profit per unit staked), p is your estimated probability of winning, and q is 1 − p. If the formula returns zero or a negative number, you have no edge and Kelly says do not bet.

Worked example: you find a bet at 2.00 (even money) you believe wins 55% of the time. Then b = 1, p = 0.55, q = 0.45, so f = (1 × 0.55 − 0.45) ÷ 1 = 0.10. Kelly suggests staking 10% of your bankroll. On a $1,000 bankroll that is $100. You can run any inputs through the Kelly criterion calculator to skip the arithmetic.

Why maximising growth is the point

Kelly is not about maximising your expected profit on any single bet — that would tell you to bet everything whenever you have an edge, which guarantees eventual ruin. Instead it maximises the growth rate of your bankroll compounded over many bets. That distinction is the whole idea: it balances growing fast against never busting, and it is provably optimal for long-run compounding.

Why serious bettors use fractional Kelly

Full Kelly is the theoretical optimum only if your probability estimates are perfect. They never are. Because Kelly is sensitive to those estimates, most disciplined bettors deliberately stake a fraction of the full amount:

Half Kelly keeps roughly three-quarters of the growth for about half the volatility — the most common choice.
Quarter Kelly is more conservative still, favoured by bettors with less certainty in their models or a lower tolerance for drawdowns.

Common mistake

Betting full Kelly on estimated (not true) probabilities. If you overrate your edge — as almost everyone does — full Kelly overstakes every bet, and the resulting swings can wipe out a real edge through sheer variance. Fractional Kelly is the cheap insurance against your own overconfidence.

Where standard Kelly breaks down

The classic formula assumes bets resolve one at a time. Real betting is messier: you often have several bets live at once, and some are correlated (same game, same player, same narrative). Staking each at full Kelly then commits far more of your bankroll than the formula intends. When bets overlap, shrink your fractions or size the whole slate as a portfolio.

Putting it into practice

Kelly is only as good as the edge you feed it, so the real work is verifying that edge exists. Track your closing line value to confirm your probability estimates are genuinely sharp, use a conservative fraction while you build that evidence, and let compounding do the rest. Size any bet in seconds with the Kelly calculator.

Frequently asked questions

Should I use full Kelly or fractional Kelly?: Almost always fractional. Full Kelly assumes your win-probability estimates are exact, which they never are. Half or quarter Kelly cuts your bankroll volatility dramatically while keeping most of the long-run growth, and it forgives the inevitable errors in your probability estimates.
What happens if I bet more than Kelly suggests?: Over-betting relative to Kelly increases growth up to a point, then sharply increases your risk of ruin. Bet at roughly double the Kelly fraction and your expected long-run growth actually falls to zero despite every bet being +EV — the variance eats the edge.
Does Kelly work if I bet several games at once?: Standard Kelly assumes one bet resolves before the next. With simultaneous or correlated bets, staking each at full Kelly over-exposes your bankroll. Scale your fractions down, or size the group as a portfolio, when bets overlap in time or outcome.
How accurate do my probabilities need to be?: Accurate enough that your edge is real, not imagined. Kelly amplifies both good and bad estimates, so a bettor who overrates their edge will systematically overstake. Sanity-check your estimates against closing line value before trusting them for stake sizing.