Strategy Center
The Kelly Criterion, Explained Honestly
Kelly maximizes long-term growth - and is more aggressive than almost any bettor can stomach. Here is the honest version.
Direct Answer
The Kelly Criterion gives the bet size that maximizes the long-run logarithmic growth of a bankroll: f = (bp - q) / b, where b is decimal odds minus one, p is the win probability, and q is 1 - p. Full Kelly is mathematically optimal but practically too volatile; most professionals use fractional Kelly.
Key Takeaways
- 01Kelly maximizes long-run geometric growth, not short-run profit.
- 02Full Kelly is aggressive enough to produce 50%+ drawdowns.
- 03Most professionals stake 25% to 50% of Kelly to reduce variance.
- 04Kelly assumes your edge estimate is exact - it rarely is.
The formula and an example
For a binary bet, Kelly fraction f = (bp - q) / b. Suppose you believe a +120 underdog wins 50% of the time. Decimal odds are 2.20, so b = 1.20, p = 0.50, q = 0.50. Kelly fraction equals (1.20 × 0.50 - 0.50) / 1.20, or 8.3% of bankroll.
Eight percent of bankroll on a single bet is enormous. A bettor with a $10,000 bankroll would risk $833 on this wager. Even with positive EV, sustained sequences of 5-10 losses are common, and drawdowns of 40-60% from peak are normal.
Why fractional Kelly wins in practice
Full Kelly maximizes long-run growth only if your probability estimate is exact. In reality, your edge is uncertain. Overestimating your edge by a small amount and betting full Kelly can convert a positive-EV strategy into a negative-EV one. Betting half Kelly recovers roughly 75% of growth with roughly 50% of the volatility.
A practical default for sports bettors: 0.25 to 0.50 Kelly on bets where edge is reasonably well-estimated; flat sizing or smaller fractions when edge is more speculative.
What Kelly will not do
Kelly does not protect against estimation error. Kelly does not account for correlated bets. Kelly does not factor in withdrawal needs or external liquidity. Treating it as a magic formula has bankrupted skilled bettors with genuine edges.
Frequently asked questions
Is Kelly always optimal?+
Kelly is optimal for maximizing the long-run geometric growth rate of a bankroll under the assumption of known, accurate edge. Any deviation - estimation error, correlated bets, liquidity constraints - argues for fractional Kelly.
What happens if I bet more than full Kelly?+
Above full Kelly, long-run growth rate falls. Past 2x Kelly, long-run growth becomes negative even with positive EV bets, because compounding volatility erodes capital faster than the edge replenishes it.
Can I use Kelly across multiple simultaneous bets?+
Yes, but the math requires accounting for correlation between outcomes. Independent bets can be sized closer to individual Kelly; correlated bets require reduced exposure to avoid concentration risk.
This article is educational only. It is not wagering, financial, or legal advice. See our editorial policy.