The Kelly Criterion tells you the bet size that maximizes long-run bankroll growth given a real edge. Enter your bankroll, your honest estimated win probability, and the offered American odds. Most pros use half or quarter Kelly to soften variance. Last updated: May 2026.
Formula: f = (b × p − q) ÷ b, where b = decimal odds − 1, p = your win probability, and q = 1 − p. If f comes out negative, you have no edge and the right bet size is zero.
This tool runs the classic Kelly Criterion formula: f* = (bp − q) ÷ b. Here b is your net decimal payout (decimal odds minus 1 — the profit you collect per $1 risked), p is the win probability you entered, and q is 1 − p, your chance of losing. The output f* is the fraction of bankroll that maximizes long-run compound growth.
Before applying Kelly, the calculator converts your American odds to decimal: positive odds become (odds ÷ 100) + 1, negative odds become (100 ÷ |odds|) + 1. It also reports the book's implied probability (1 ÷ decimal odds) and your edge (your probability minus that implied figure). If your edge is zero or negative, f* drops to zero or below — and the correct stake is nothing. The half and quarter Kelly numbers are just f* multiplied by 0.5 and 0.25, the fractional approach most disciplined bettors actually use to tame variance.
Take the default inputs: a $1,000 bankroll, a 55% win estimate, and odds of −110. First, −110 converts to decimal: (100 ÷ 110) + 1 = 1.909, so b = 0.909. With p = 0.55 and q = 0.45, the math is f* = (0.909 × 0.55 − 0.45) ÷ 0.909 = (0.50 − 0.45) ÷ 0.909 = 0.055, or 5.5% of bankroll.
Full Kelly therefore recommends 0.055 × $1,000 = $55.00. Half Kelly cuts that to $27.50 and quarter Kelly to $13.75 — exactly the figures the calculator displays. The book's implied probability at −110 is 1 ÷ 1.909 = 52.38%, so your stated edge is 55% − 52.38% = 2.62 percentage points. Notice how thin that edge is: trimming your win estimate to 53% would shrink the recommended stake to roughly $13, and dropping below 52.38% would zero it out entirely. That brutal sensitivity is the whole reason to size cautiously.
I treat Kelly as a ceiling, not a target. When my edge estimate is uncertain — which is most of the time — I bet a fraction of what the formula suggests and never chase a number I can't honestly defend.
A formula developed by John Kelly in 1956 that gives the bet fraction maximizing long-run logarithmic growth of a bankroll. It assumes you have a known edge and can rebet repeatedly.
Full Kelly is mathematically optimal but extremely volatile — a string of losses can cut your bankroll in half. Half Kelly captures roughly 75% of the growth with much lower variance, which is why most professional bettors use it.
Kelly is brutally sensitive to bad inputs. Overestimating your edge by even a small amount can flip the recommendation from "bet" to "ruin." When in doubt, use a smaller fraction of Kelly or skip the bet.
It means the bet is -EV — your estimated win probability is lower than the implied probability of the odds. Kelly's correct recommendation is zero. Don't bet.
The basic single-bet Kelly formula assumes one wager at a time. For simultaneous bets, you need fractional Kelly across the slate or a more advanced "Kelly portfolio" model.