What Is the Kelly Criterion?
The Kelly Criterion is a mathematical formula that calculates the optimal percentage of your bankroll to wager on a bet with a known edge. Developed by John L. Kelly Jr. at Bell Labs in 1956, it was originally designed for information theory but has since become the gold standard for bet sizing among professional gamblers, sports bettors, and financial traders.
The core idea is straightforward: bet more when your edge is larger, bet less when your edge is smaller, and never bet when you have no edge. Kelly maximizes the long-term growth rate of your bankroll while minimizing the risk of ruin — at least in theory.
The Kelly Criterion Formula
The standard Kelly formula for sports betting is:
f* = (bp − q) / b
Where:
| Variable | Meaning |
|---|---|
| f* | The fraction of your bankroll to wager |
| b | The decimal odds minus 1 (i.e., the net payout per dollar wagered) |
| p | The probability of winning (your estimated true probability) |
| q | The probability of losing (1 − p) |
Understanding the Variables
- b converts your sportsbook's odds into a "profit per unit" figure. If the decimal odds are 2.50, then b = 1.50 (you profit $1.50 for every $1 wagered).
- p is the true probability of the bet winning — which you derive from devigging sharp lines (like Pinnacle or Circa Sports).
- q is simply the complement: if p = 0.55, then q = 0.45.
Worked Example: Moderate Edge
You've identified a +EV NFL spread bet:
| Component | Value |
|---|---|
| Sportsbook odds | +130 (decimal 2.30) |
| Fair probability of winning (devigged) | 48% |
| b (decimal odds − 1) | 1.30 |
| p | 0.48 |
| q | 0.52 |
f* = (1.30 × 0.48 − 0.52) / 1.30 f* = (0.624 − 0.52) / 1.30 f* = 0.104 / 1.30 f* = 0.08 → 8.0% of bankroll
Kelly says to wager 8% of your bankroll on this bet.
Worked Example: Small Edge
Now consider a bet with a smaller edge:
| Component | Value |
|---|---|
| Sportsbook odds | −110 (decimal 1.909) |
| Fair probability of winning | 54% |
| b | 0.909 |
| p | 0.54 |
| q | 0.46 |
f* = (0.909 × 0.54 − 0.46) / 0.909 f* = (0.491 − 0.46) / 0.909 f* = 0.031 / 0.909 f* = 0.034 → 3.4% of bankroll
A smaller edge yields a smaller recommended stake — exactly the behavior you want from a sizing formula.
Worked Example: Large Edge
Consider a player prop with a significant mispricing:
| Component | Value |
|---|---|
| Sportsbook odds | +180 (decimal 2.80) |
| Fair probability of winning | 45% |
| b | 1.80 |
| p | 0.45 |
| q | 0.55 |
f* = (1.80 × 0.45 − 0.55) / 1.80 f* = (0.81 − 0.55) / 1.80 f* = 0.26 / 1.80 f* = 0.144 → 14.4% of bankroll
Full Kelly recommends nearly 15% of your bankroll on this single bet. That brings us to a critical point.
Why Full Kelly Is Too Aggressive in Practice
The Kelly Criterion is mathematically optimal under very specific assumptions: you know the exact true probability, you're placing one bet at a time, and you have an infinite time horizon. In reality, none of these hold perfectly.
The Problems with Full Kelly
- Probability estimation error. Your devigged fair probability is an estimate, not a certainty. If you overestimate your edge by even a small amount, full Kelly can lead to catastrophic overbetting.
- Bankroll volatility. Full Kelly produces extreme swings. Simulations show that full Kelly bettors frequently experience 50–80% drawdowns before recovering — most people can't stomach that.
- Simultaneous bets. Kelly assumes one bet at a time. If you're placing 10 bets on an NFL Sunday, full Kelly on each would overexpose your bankroll.
Full Kelly maximizes growth rate but tolerates massive variance. For real-world sports betting, the solution is fractional Kelly.
Fractional Kelly: The Practical Approach
Fractional Kelly means multiplying the full Kelly stake by a fraction — typically 0.25 (quarter Kelly) or 0.50 (half Kelly).
| Kelly Variant | Formula | 8% Full Kelly Becomes |
|---|---|---|
| Full Kelly | f* | 8.0% |
| Half Kelly | f* × 0.5 | 4.0% |
| Quarter Kelly | f* × 0.25 | 2.0% |
Which Fraction Should You Use?
- Quarter Kelly (0.25×) is the most conservative and most commonly recommended starting point. It dramatically reduces variance while still capturing most of the long-term growth.
- Half Kelly (0.50×) balances growth and volatility well for bettors with accurate models and strong conviction in their edge estimates.
- Full Kelly (1.0×) is almost never used in practice by experienced bettors. The theoretical growth-rate advantage over half Kelly is small, while the additional variance is severe.
Most professional sports bettors and betting syndicates operate in the quarter-to-half Kelly range.
Handling Simultaneous Bets
On a typical NFL Sunday, you might identify 8–12 +EV bets. If you apply Kelly independently to each, the total bankroll at risk could exceed 30–40%, which is far too aggressive.
Approaches for simultaneous bets:
- Reduce to quarter Kelly or lower to account for the total exposure across all concurrent wagers.
- Cap total exposure at a fixed percentage of your bankroll (e.g., 15–20% total across all bets in a day).
- Prioritize the highest-EV bets and reduce sizing on marginal edges.
The exact solution involves multivariate Kelly, which accounts for correlations between bets — but for most bettors, simply using quarter Kelly and capping total daily exposure is practical and effective.
Kelly Criterion vs. Flat Betting
Flat betting means wagering the same fixed amount (e.g., 2% of your bankroll) on every bet regardless of edge size. How does it compare to Kelly?
| Factor | Kelly | Flat Betting |
|---|---|---|
| Growth rate | Higher (mathematically optimal) | Lower |
| Simplicity | Requires edge estimate per bet | Simple — same stake every time |
| Variance | Higher (even fractional Kelly) | Lower and more predictable |
| Sensitivity to errors | Punishes overestimated edges | Unaffected by edge estimation |
| Best for | Bettors with reliable edge estimates | Bettors who want simplicity |
If your +EV model is well-calibrated and you trust your fair probability estimates, fractional Kelly will grow your bankroll faster. If you're less confident in your estimates or just starting out, flat betting at 1–2% per wager is a perfectly reasonable approach that still captures your edge.
Common Mistakes with the Kelly Criterion
Overestimating Your Edge
The single most dangerous mistake. If you think a bet has a 5% edge but it actually has a 1% edge, full Kelly will have you wagering far too much. Always be conservative with your probability estimates, and use fractional Kelly as an additional safety buffer.
Ignoring Bankroll Updates
Kelly sizing should be based on your current bankroll, not your starting bankroll. If your bankroll grows from $5,000 to $7,000, your bet sizes should increase proportionally. If it drops to $3,500, your bet sizes should decrease. This automatic scaling is one of Kelly's greatest strengths — it naturally reduces your exposure during losing streaks.
Applying Kelly to −EV Bets
The Kelly formula returns a negative number when the bet has negative expected value. That's the formula telling you not to bet at all. Never override this — if Kelly says don't bet, there's no edge to exploit.
Using Full Kelly on Parlays or Longshots
Longshot bets (high decimal odds, low win probability) produce enormous Kelly fractions that can wipe out your bankroll on a single loss. Fractional Kelly is especially important for these wagers.
How WagerWiz Calculates Recommended Bet Sizes
WagerWiz integrates Kelly-based staking directly into its +EV screener. For every flagged bet, the platform displays:
- The full Kelly percentage based on the devigged fair probability and the sportsbook's offered odds
- A recommended fractional Kelly stake (configurable — quarter, third, or half Kelly)
- The dollar amount based on the bankroll size you set in your account
This means you don't need to plug numbers into a spreadsheet for every bet. You set your bankroll, choose your Kelly fraction, and WagerWiz tells you exactly how much to wager on each opportunity.
FAQ
Is the Kelly Criterion the same as the Kelly formula?
Yes. "Kelly Criterion," "Kelly formula," "Kelly strategy," and "Kelly staking" all refer to the same concept: using f* = (bp − q) / b to determine optimal bet sizing based on your edge and the odds offered.
What Kelly fraction do professional sports bettors use?
Most professionals use between quarter Kelly (0.25×) and half Kelly (0.50×). The exact fraction depends on their confidence in their probability model, the number of simultaneous bets they place, and their personal risk tolerance. Very few experienced bettors use full Kelly.
Can I use the Kelly Criterion for parlays?
In theory, yes — the formula works for any bet with a known edge and known odds. In practice, parlays are tricky because estimating the true probability of a multi-leg parlay requires accounting for correlations between legs, which introduces additional estimation error. Use fractional Kelly and be very conservative with parlay sizing.
What happens if the Kelly formula gives a negative number?
A negative result means the bet has negative expected value — the sportsbook's odds don't offer enough compensation for the risk. Kelly is telling you not to bet. Always listen to the math.
How does Kelly interact with bankroll management rules like "never bet more than 5%"?
They complement each other. The Kelly Criterion tells you the mathematically optimal stake, and a hard cap (like 5% of bankroll) acts as an additional guardrail. If full Kelly says to bet 14% but your rule says max 5%, bet 5%. In practice, if you're using quarter Kelly, your stakes will almost always fall below 5% naturally.
