How much should you bet? It’s the most important question in sports betting—and most bettors get it completely wrong.
The Kelly Criterion is a mathematical formula that tells you exactly how much to stake based on your edge and the odds. Used correctly, it maximizes long-term bankroll growth. Used incorrectly—or ignored entirely—it’s why most bettors eventually go broke.
What Is the Kelly Criterion?
The Kelly Criterion is a formula developed by John Kelly at Bell Labs in 1956. Originally designed for information theory, it was quickly adopted by gamblers and investors for optimal bet sizing.
The Core Concept
Kelly answers a simple question: Given my edge and the odds, what percentage of my bankroll should I bet to maximize long-term growth?
The answer isn’t intuitive. Bet too little, and you’re leaving money on the table. Bet too much, and variance will eventually wipe you out—even with a genuine edge.
The Kelly Criterion Formula
Standard Formula
f* = (bp – q) / b
Where:
- f* = Fraction of bankroll to bet
- b = Decimal odds – 1 (net odds received on win)
- p = Probability of winning
- q = Probability of losing (1 – p)
Example Calculation
You find a bet at +150 (2.50 decimal) where you estimate 45% true win probability.
- b = 2.50 – 1 = 1.50
- p = 0.45
- q = 0.55
f* = (1.50 × 0.45 – 0.55) / 1.50
f* = (0.675 – 0.55) / 1.50
f* = 0.125 / 1.50
f* = 0.0833 = 8.33%
Kelly says bet 8.33% of your bankroll on this bet.
Simplified Formula for American Odds
For American odds, use:
For positive odds (+150):
f* = (p × (odds/100 + 1) – 1) / (odds/100)
For negative odds (-150):
f* = (p × (100/|odds| + 1) – 1) / (100/|odds|)
Why Kelly Works
Maximizes Geometric Growth
Kelly optimizes for geometric growth, not arithmetic. This matters because betting is multiplicative—a 50% loss requires a 100% gain to recover.
Example over 10 bets with a 55% win rate at -110:
- Flat betting 10% per bet: Volatile, possible ruin
- Kelly betting (~2.3%): Steady growth, minimal ruin risk
Accounts for Variance
Bigger edges = bigger Kelly stakes. Smaller edges = smaller stakes. The formula automatically adjusts for the risk/reward of each situation.
Prevents Overbetting
The biggest bankroll killer is overbetting. Kelly mathematically prevents this by capping your stake based on actual edge.
The Problem With Full Kelly
Here’s why almost no serious bettor uses full Kelly: the stakes are too aggressive.
Volatility Is Extreme
Full Kelly optimizes for maximum long-term growth, which means accepting massive short-term swings. A 50% drawdown is common—and psychologically brutal.
Edge Estimation Is Imperfect
Kelly assumes you know your exact edge. You don’t. If you overestimate your edge by even a small amount, full Kelly will destroy your bankroll.
Real-World Example
If Kelly says bet 8% and your actual edge is half what you estimated, you’re massively overbetting. The formula becomes a liability.
Fractional Kelly: The Practical Solution
Most sharp bettors use fractional Kelly—typically 25-50% of the full Kelly stake.
Why Fractional Kelly Works Better
- Reduces volatility: Half Kelly cuts variance by ~75%
- Buffers estimation error: If your edge estimate is off, you’re protected
- Psychologically manageable: Smaller swings = better decision-making
- Still compounds: Growth is slower but steadier
Fractional Kelly Recommendations
| Experience Level | Recommended Fraction | Why |
|---|---|---|
| Beginner | 25% Kelly | Large buffer for estimation errors |
| Intermediate | 33-50% Kelly | Balanced growth and protection |
| Advanced | 50-75% Kelly | Confident in edge estimation |
| Professional | Varies by bet type | Sophisticated modeling |
Kelly Criterion in Practice
Step 1: Estimate Your Edge
This is the hard part. You need to know your true win probability—not what the odds imply, but what you believe it actually is.
Methods for estimating edge:
- Model-based: Build or use a predictive model
- Closing line value (CLV): Track how often you beat the closing line
- Historical results: Your long-term win rate (1,000+ bets minimum)
- Sharp consensus: Compare to sharp book numbers
For more on finding edge, see our guide on reading betting action.
Step 2: Calculate Kelly Stake
Use the formula or a calculator:
- OddsJam Kelly Calculator
- SBR Kelly Calculator
- Spreadsheet formula (see below)
Step 3: Apply Fractional Adjustment
Multiply full Kelly by your chosen fraction (e.g., 0.5 for half Kelly).
Step 4: Check Against Maximum
Even with Kelly, cap your bets. Most professionals never exceed 3-5% on any single bet, regardless of what Kelly suggests.
Excel/Sheets Kelly Formula
For decimal odds in cell A1 and win probability in B1:
=((B1*(A1-1)-(1-B1))/(A1-1))
For half Kelly, multiply by 0.5:
=((B1*(A1-1)-(1-B1))/(A1-1))*0.5
When Kelly Doesn’t Apply
Parlays
Kelly assumes independent bets. Parlays are correlated—use a different approach or avoid Kelly sizing entirely for parlays.
Simultaneous Bets
Standard Kelly is for sequential bets. If you’re betting multiple games at once, you need to adjust for the total risk exposure.
Correlated Outcomes
If your bets are correlated (same game, related props), simple Kelly overcounts your edge. Adjust accordingly.
Limited Bankroll
If your bankroll is small relative to minimum bet sizes, Kelly math breaks down. Focus on building bankroll first.
Kelly vs. Flat Betting vs. Unit Systems
Flat Betting (Fixed Stakes)
Pros: Simple, consistent, easy to track
Cons: Doesn’t adjust for edge size, suboptimal growth
Unit Systems (1-5 Units)
Pros: Allows confidence-based sizing
Cons: Arbitrary, no mathematical basis
Kelly
Pros: Mathematically optimal, accounts for edge
Cons: Requires accurate edge estimation, complex
Recommendation
If you can estimate your edge accurately, use fractional Kelly. If you can’t, use flat betting with 1-2% stakes until you can.
For more on bankroll management strategies, see our complete bankroll management guide.
Common Kelly Mistakes
Mistake 1: Overestimating Edge
The #1 Kelly killer. If you think you have 5% edge but actually have 2%, you’re overbetting by 150%.
Solution: Be conservative with edge estimates. Use fractional Kelly as a buffer.
Mistake 2: Using Full Kelly
Full Kelly variance is not for humans. The math works; your psychology won’t.
Solution: Start at 25-33% Kelly maximum.
Mistake 3: Ignoring Correlation
Betting Chiefs -3 and Chiefs over 45 on the same game? Those outcomes are correlated. Kelly treats them as independent.
Solution: Reduce stakes on correlated bets or treat them as a single position.
Mistake 4: Recalculating Too Frequently
Recalculating Kelly after every bet creates transaction costs and psychological stress.
Solution: Recalculate weekly or after significant bankroll changes (±20%).
Kelly for Different Bet Types
Spread Bets (-110)
At -110, break-even is 52.4%. A 55% bettor has ~2.5% edge.
Full Kelly: ~2.3% of bankroll
Half Kelly: ~1.15% of bankroll
Moneylines (Favorites)
Heavy favorites have low Kelly stakes because the edge per dollar risked is small, even with high win rates.
Moneylines (Underdogs)
Underdogs with edge can have high Kelly stakes. This is where fractional Kelly is crucial—a +300 with perceived 30% win probability suggests 5% stake, which is aggressive.
Props
Props often have larger edges but higher variance. Use more conservative fractional Kelly (25-33%).
Building a Kelly System
Requirements
- Edge estimation method: Model, CLV tracking, or sharp line comparison
- Bankroll tracking: Know your current bankroll precisely
- Calculator or spreadsheet: Automate the math
- Discipline: Actually follow the stakes Kelly recommends
Sample Workflow
- Identify potential bet
- Estimate true win probability
- Calculate implied probability from odds
- If edge exists (true > implied), calculate Kelly stake
- Apply fractional adjustment
- Check against maximum stake limit
- Place bet
Final Thoughts
The Kelly Criterion is the mathematically optimal way to size bets—but only if you know your edge. For most bettors, the practical takeaways are:
- Never use full Kelly — too volatile
- Start with 25-33% Kelly — build confidence
- Cap maximum stakes at 3-5% — regardless of Kelly output
- Be conservative with edge estimates — overconfidence kills
- If you can’t estimate edge, use flat betting — 1-2% stakes
Kelly won’t make bad bets good. But for bets where you have genuine edge, it ensures you’re staking the right amount to maximize long-term growth without blowing up.
For more on finding edges to bet, see our guides on betting tools and implied probability.
