↓ Skip to main content

Kelly Criterion Betting Formula, With Fractional Kelly Examples

·5 mins

The Kelly criterion tells you what fraction of your bankroll to stake on a bet with a positive edge, so that your bankroll grows at the fastest possible long-run rate. The formula:

f* = (b × p − q) / b

  • f* = fraction of bankroll to bet
  • b = net odds (profit per $1 staked), which is decimal odds − 1
  • p = your probability of winning
  • q = 1 − p

An equivalent form that is easy to remember: f* = edge / net odds, where edge = p × decimal odds − 1.

If f* comes out zero or negative, the bet has no edge and Kelly says don’t bet. The catch, covered at length below, is that p is an estimate, and Kelly punishes overestimates hard.

Odds here are illustrative examples.

Where the formula comes from

J. L. Kelly Jr. published it in the Bell System Technical Journal in 1956 as “A New Interpretation of Information Rate”. His paper (available as a PDF reproduced with AT&T’s permission) shows that a gambler with an information advantage can make capital “grow exponentially”, and that the maximum growth rate is tied to the information rate of the channel. The betting rule falls out of maximizing the expected logarithm of wealth.

Worked example 1: an underdog at +150

You estimate a 45% win chance. +150 is decimal 2.50, so b = 1.50.

  1. b × p = 1.50 × 0.45 = 0.675
  2. q = 0.55
  3. f* = (0.675 − 0.55) / 1.50 = 0.125 / 1.50 = 0.0833, so 8.33% of bankroll

Check with the edge form: edge = 0.45 × 2.50 − 1 = 0.125. Divide by b: 0.125 / 1.50 = 0.0833. Same answer.

Worked example 2: a spread bet at -110

You estimate 55%. -110 is decimal 1.909, so b = 0.909.

  • f* = (0.909 × 0.55 − 0.45) / 0.909 = (0.500 − 0.45) / 0.909 = 5.50%
  • At 53%: f* = 1.30%
  • At 52.38% (the break-even rate at -110): f* = 0%

Small changes in p move the stake a lot. That sensitivity is the whole problem with full Kelly.

Fractional Kelly tables

Most bettors who use Kelly bet a fraction of it. On a $5,000 bankroll:

Bet Full Kelly Half Kelly Quarter Kelly
+150, p = 45% 8.33% ($417) 4.17% ($208) 2.08% ($104)
-110, p = 55% 5.50% ($275) 2.75% ($138) 1.38% ($69)

Growth rate: why less can be better

Expected log growth per bet is G(f) = p × ln(1 + b × f) + q × ln(1 − f). For the +150 example (p = 0.45, b = 1.5):

Stake Share of bankroll Growth per bet Share of full-Kelly growth
Quarter Kelly 2.08% 0.227% 44%
Half Kelly 4.17% 0.387% 75%
Full Kelly 8.33% 0.515% 100%
1.5× Kelly 12.50% 0.389% 76%
2× Kelly 16.67% 0.014% 3%
2.5× Kelly 20.83% −0.612% negative

Three takeaways:

  • Half Kelly keeps about 75% of the growth of full Kelly with half the stake size.
  • Overbetting is worse than underbetting. 1.5× Kelly grows at about the same rate as half Kelly while staking three times as much.
  • At about 2× Kelly, growth falls to roughly zero, and beyond it your bankroll is expected to shrink even though every bet has a positive edge.

Drawdowns: a simulation

Same bet (+150, true p = 45%), 1,000 independent bets, 20,000 simulated runs each (our own simulation, assuming p is known exactly):

Staking Bankroll fell to half its start at some point Ended below starting bankroll
Quarter Kelly 0.7% 0.3%
Half Kelly 12.1% 0.8%
Full Kelly 47.6% 5.3%
2× Kelly 90.0% 48.5%

Full Kelly halved the bankroll at some point in almost half the runs, even with a perfectly known edge. Real bettors don’t know their edge that precisely.

What happens when your p is wrong

Say you think p = 45% at +150, but the true chance is 42%.

  • Your Kelly stake: 8.33%.
  • The correct Kelly stake at 42%: (1.5 × 0.42 − 0.58) / 1.5 = 3.33%.
  • Your growth per bet at 8.33% with the true 42%: −0.10%. You are overbetting by 2.5× and shrinking your bankroll on a bet that is still +5% EV.
  • Half Kelly (4.17%) at the true 42%: +0.08% per bet. Still growing.

This is the practical case for fractional Kelly: it gives you a margin of safety against overconfident probability estimates. The fair-probability methods in no-vig odds and the edge math in expected value betting explained are where your p comes from.

Calculator steps

No embedded calculator here, but the spreadsheet is short. With American odds in A2 and your probability in B2:

  1. Decimal odds (C2): =IF(A2>0,1+A2/100,1+100/ABS(A2))
  2. Net odds b (D2): =C2-1
  3. Kelly fraction (E2): =(D2*B2-(1-B2))/D2
  4. Stake (F2): =MAX(0,E2)*bankroll*fraction, with fraction set to 1, 0.5 or 0.25

Tools we reference Several +EV tools output a Kelly or fractional-Kelly stake alongside each flagged bet. We compare their published features and prices in best EV betting tools. A tool’s Kelly number is only as good as the fair probability it plugs in.

Practical limits

  • Simultaneous bets. The single-bet formula assumes one bet at a time. Running several full-Kelly bets at once overstakes; scale down.
  • Correlated bets. Two bets on the same game aren’t independent. Treat them as one position.
  • Max bets. Licensed books can cap your stake regardless of what Kelly says; DraftKings’ Massachusetts house rules let it lower any customer’s maximum payout at its discretion (implemented Aug 26, 2025, checked Nov 5, 2026).
  • Bankroll definition. Kelly uses your dedicated betting bankroll, not your savings. See bankroll management.

Betting is restricted to adults (21+ in most states). If staking plans start to feel like chasing, call or text 1-800-MY-RESET, and read our responsible gambling page.