The Kelly Criterion is a mathematical formula that tells you the optimal percentage of your bankroll to bet based on your edge. Used correctly, it maximizes long-term growth. Used incorrectly, it can blow up your bankroll. Here's how it works.
The Kelly Formula
f* = fraction of bankroll to bet
b = decimal odds - 1 (or American odds converted)
p = probability of winning
q = probability of losing (1 - p)
A Real Example
Let's say you believe a bet at +110 odds has a 55% chance of winning. Here's how to calculate the Kelly stake:
Example: +110 with 55% Win Probability
Kelly says to bet 4.5% of your bankroll on this wager. That's aggressive—which is why most professionals use fractional Kelly.
Why Full Kelly Is Dangerous
The Kelly Criterion assumes you know your exact edge. In sports betting, you never do. Your "55% probability" is an estimate, not a certainty.
Problems with full Kelly:
- Overestimated edges: If you think you have a 55% edge but really have 52%, Kelly overestimates your bet size
- Variance: Full Kelly creates massive swings—50%+ drawdowns are common
- Emotional tolerance: Most humans can't stomach Kelly-sized losses
- Practical limits: Sportsbook limits often prevent you from betting Kelly-recommended amounts anyway
Kelly requires knowing your true win probability. But in sports betting, you're always estimating. A slight overestimate of your edge leads to dramatically overbetting. This is why full Kelly often leads to ruin.
Fractional Kelly: The Practical Approach
Most successful bettors use fractional Kelly—typically 25% to 50% of the full Kelly recommendation:
| Kelly Fraction | Bet Size (from 4.5% example) | Trade-off |
|---|---|---|
| Full Kelly (100%) | 4.5% | Maximum growth, maximum variance |
| Half Kelly (50%) | 2.25% | Good growth, manageable variance |
| Quarter Kelly (25%) | 1.1% | Slower growth, smooth ride |
| Flat 1% | 1% | Simple, ignores edge sizing |
Half Kelly is a common choice among professionals. It captures most of the growth potential while dramatically reducing variance. Quarter Kelly is even more conservative and approximates the flat 1% rule we recommend for most bettors.
When Kelly Makes Sense
Kelly is most useful when you have:
- A proven edge: Tracked over 500+ bets with positive CLV
- Quantifiable probability estimates: A model, not gut feel
- Patience: Ability to weather drawdowns without panicking
- Multiple simultaneous bets: Kelly works best when diversifying across many edges
If you're new to betting or don't have a tracked edge, Kelly is premature. Stick with flat staking until you've proven yourself profitable over a large sample.
Kelly vs. Flat Betting
| Factor | Kelly | Flat 1% |
|---|---|---|
| Requires edge estimate | Yes (can be wrong) | No |
| Complexity | High | Simple |
| Long-term growth | Optimal (if edge is known) | Sub-optimal but steady |
| Variance | High | Low |
| Ruin probability | Higher if edge overestimated | Very low |
Even sharp bettors with proven models often use quarter Kelly or less. The theoretical maximum growth isn't worth the emotional and practical costs of full Kelly variance. Survival matters more than optimization.
Common Kelly Mistakes
1. Using Kelly Without a Proven Edge
If you don't have a tracked, statistically significant edge, Kelly gives you nothing. Garbage in, garbage out.
2. Ignoring Correlation
Kelly assumes independent bets. If you're betting multiple correlated outcomes (same game, related markets), you need to adjust down.
3. Not Adjusting for Limits
Kelly might say bet $500, but the book limits you to $200. Real-world constraints matter.
4. Emotional Sabotage
After a 30% drawdown, most bettors panic and abandon their system. If you can't handle the swings, use smaller stakes.
Our Recommendation
For most Ohio bettors, the 1% flat staking rule is superior to Kelly. It's simple, protects your bankroll, and doesn't require accurate edge estimation.
Consider fractional Kelly (quarter to half) only if:
- You have 500+ tracked bets with positive CLV
- You use a quantitative model for probability estimation
- You can emotionally handle significant drawdowns
- You understand you might still be wrong
The best system is the one you'll actually follow. For most people, that's flat staking.