Introduction
The difference between profitable bettors and losing punters often comes down to one critical factor: stake sizing. You can identify value in the market, build a data-driven selection process, and maintain discipline — but if you bet too much on any single outcome, variance will eventually destroy your bankroll. Bet too little, and you leave profit on the table.
This is where the Kelly Criterion comes in.
Originally developed in 1956 by John Kelly Jr., a researcher at Bell Labs working on telephone signal noise, the Kelly Criterion is a mathematical formula that calculates the optimal bet size to maximize long-term bankroll growth while minimizing the risk of ruin. What began as a solution to telecommunications problems quickly became recognized by gamblers and investors as the gold standard for bankroll management.
The Kelly Criterion answers a simple question: How much should I stake when I have an edge?
In this guide, we'll break down the mathematics behind Kelly, explain how to apply it to football player props and tennis betting, address its limitations, and show you why most professional bettors use a modified version called "fractional Kelly" rather than the full formula.
The Mathematics Explained
The Kelly Formula
The Kelly Criterion calculates the optimal fraction of your bankroll to wager using this formula:
f* = (bp - q) / b
Where:
- f* = the fraction of your bankroll to bet (the Kelly percentage)
- b = the net odds received on the wager (decimal odds minus 1)
- p = your estimated probability of winning
- q = probability of losing (1 - p)
Breaking It Down With an Example
Let's say you're betting on a player props market:
- Your assessment: A player has a 55% chance of 2+ shots on target
- Bookmaker odds: 1.91 (decimal) or -110 (American)
- Your bankroll: £1,000
First, calculate the variables:
- b = 1.91 - 1 = 0.91
- p = 0.55 (55%)
- q = 0.45 (45%)
Now apply the formula:
f* = (0.91 × 0.55 - 0.45) / 0.91
f* = (0.5005 - 0.45) / 0.91
f* = 0.0505 / 0.91
f* = 0.055 or 5.5%
The Kelly Criterion recommends staking 5.5% of your bankroll, which equals £55 on this bet.
Why This Works
The Kelly Criterion maximizes the geometric growth rate of your bankroll over time. Unlike arithmetic approaches (flat staking), Kelly accounts for the compounding effect of wins and the devastating impact of losses.
The formula effectively balances two competing forces:
- Betting enough to capitalize on your edge
- Betting conservatively enough to survive inevitable losing runs
Crucially, if the Kelly formula returns zero or a negative number, it tells you not to bet — the odds don't justify the risk. This built-in safeguard prevents you from placing wagers where you don't have a genuine edge.
Understanding the Relationship
Key insight: Kelly bet size increases with:
- Larger edges (higher p relative to implied probability)
- Higher odds (more b)
Kelly bet size decreases with:
- Smaller edges
- Lower odds (less b)
This makes intuitive sense: you should bet more when you have a significant advantage at favorable odds, and less when your edge is marginal.
Kelly Criterion Quick Summary
- • Calculates optimal bet size based on your edge
- • Formula:
f* = (bp - q) / b - • Use fractional Kelly (one-tenth for props, quarter for tennis)
- • Never bet more than 5% of bankroll on single bet
- • Only works if you have a genuine edge
Kelly in Practice — Player Props & Tennis
Applying Kelly to Football Player Props
Player props markets — shots on target, fouls committed, tackles won — are where bookmakers often misprice individual player performance. Here's how Kelly applies:
Example: Player Props Calculation
Scenario: Declan Rice 2+ tackles won vs Aston Villa
- • Rice averages 2.3 tackles per game in the Premier League
- • Villa's midfield tends to give up tackle opportunities
- • Your matchup analysis suggests 58% probability of 2+ tackles won
- • Bookmaker offers 2.00 (decimal)
Kelly calculation:
b = 1.00, p = 0.58, q = 0.42 → f* = 16%
Full Kelly suggests staking 16% of your bankroll. On £1,000 that's £160 on a single prop bet. One bad result and you've lost a big chunk. Player props are more volatile than match outcomes (individual performance swings harder), and probability estimates are less reliable than in liquid tennis markets. That's too aggressive. This is why many professionals use fractional Kelly, often one-tenth or quarter Kelly for props.
Fractional Kelly: The Professional Approach
Rather than betting the full Kelly percentage, most sharp bettors use a fraction:
- One-Tenth Kelly (0.1x): Bet 10% of the Kelly recommendation. Advised for prop bets because they are more volatile than match outcomes: individual player performance swings harder, and probability estimates are less reliable.
- Quarter Kelly (0.25x): Bet 25% of the Kelly recommendation
- Half Kelly (0.5x): Bet 50% of the Kelly recommendation
Using our Rice example with one-tenth Kelly:
Full Kelly: 16% → One-tenth Kelly: 16% × 0.1 = 1.6% of bankroll
On a £1,000 bankroll, that's £16 instead of £160.
Why Fractional Kelly?
Three critical reasons:
1. Probability Estimation Error — You're never 100% certain of your assessed probability. Fractional Kelly provides a safety buffer.
2. Variance Tolerance — Full Kelly produces violent bankroll swings. Half Kelly reduces volatility dramatically while sacrificing only 25% of long-term growth rate.
3. Multiple Simultaneous Bets — If you're betting 20-30 props per week, full Kelly on each bet would have you massively overexposed. Fractional Kelly allows you to spread risk across a portfolio of bets.
Applying Kelly to ATP Tennis
Tennis markets operate similarly, but with different characteristics:
Example: ATP Tennis
Scenario: ATP 250 first set winner
- • Your model: Player A has 58% chance to win first set
- • Pinnacle (closing line): 1.85 (implies 54% probability)
- • You've found a 4% edge
Kelly: f* = 6.95%. With quarter Kelly: 1.74%
On £2,000 bankroll: £34.80 stake
This aligns with our staking range of 0.5u to 2u (where 1 unit = 1-2% of bankroll).
Kelly Across Different Markets
The beauty of Kelly is it automatically scales your stake based on:
- Edge size: Bigger edge = larger stake
- Odds: Kelly accounts for risk/reward relationship
- Bankroll: Stakes grow as bankroll grows, shrink as it shrinks
This dynamic sizing is far superior to flat staking, which treats all bets identically regardless of edge or context.
The Critical Limitations
The Kelly Criterion Is Not Perfect
Before you rush to implement Kelly, understand its weaknesses. Professional bettors know these limitations intimately:
1. Extreme Sensitivity to Probability Errors
The Kelly Criterion's greatest flaw: it assumes you know the true probability. You don't. You never do.
Real-world example: You estimate 60% win probability, but true probability is 55%. Full Kelly recommends 8.2% of bankroll; correct Kelly (if you knew true odds) is 2.1%. You're betting nearly 4x too much due to a 5-point estimation error.
The mathematics are unforgiving: if you bet twice the optimal Kelly amount, your long-term growth rate becomes zero. Bet more than 2x Kelly, and you'll lose money over time despite having an edge.
⚠️ The Danger of Overconfidence
If you estimate 60% win probability but true probability is 55%, you'll bet nearly 4x too much. This is why professionals use fractional Kelly (one-tenth or quarter) — it provides a safety margin against estimation errors.
2. Assumes Accurate Odds Assessment
Kelly requires you to accurately quantify your edge. For player props, this means collecting season-long data, adjusting for opponent tendencies, accounting for injuries, motivation, form, and understanding bookmaker pricing patterns. If your analysis is flawed, Kelly will magnify those errors.
3. Bankroll Volatility
Full Kelly: 50% chance of 50% drawdown, 10% chance of 75% drawdown. Most bettors cannot psychologically handle this. Quarter Kelly reduces these swings dramatically.
4. Transaction Costs & Liquidity
Kelly assumes zero transaction costs, infinite liquidity, ability to bet any amount. Real-world: bookmakers limit stakes, you face restrictions after winning, minimum bet sizes exist.
5. The Multiple Bets Problem
Kelly was designed for sequential, single bets. But profitable strategies generate multiple simultaneous opportunities. If you're betting 30 props across a weekend, each at 2% Kelly, you're effectively 60% leveraged on correlated outcomes.
6. Requires Discipline
Kelly tells you to bet more after wins and less after losses. Most bettors do the opposite. Kelly demands you ignore emotional impulses.
7. No Edge? Don't Bet
If Kelly returns zero or negative, you don't have an edge — regardless of how confident you feel. The formula is brutally honest.
When Kelly Fails Completely
Kelly assumes: genuine edge, accurate probability estimates, sufficient bankroll, ability to handle volatility, long-term perspective (hundreds of bets minimum). If any aren't met, Kelly is the wrong tool. Recreational bettors with small bankrolls should use fixed small stakes instead.
Fractional Kelly — Which Multiple to Use
Professional bettors rarely use full Kelly. The question isn't whether to use a fraction, but which fraction.
| Kelly Fraction | Volatility | Growth Rate | Best For |
|---|---|---|---|
| Full Kelly (1x) | Extreme | 100% | Almost nobody |
| Half Kelly (0.5x) | High | 75% | Aggressive pros |
| Quarter Kelly (0.25x) | Moderate | 56% | Tennis / match markets (liquid) |
| One-Tenth Kelly (0.1x) | Low | 19% | Prop bets / conservative (most pros) |
How Il Margine Approaches It
We use one-tenth Kelly for prop bets — they're more volatile and our probability estimates are less reliable. For tennis (match / first set markets), where edges are easier to quantify and markets are more liquid, we use quarter Kelly.
Our stake range is generally 0.5u to 2u, where 1 unit ≈ 1% of bankroll. That means:
- • Minimum: ~0.5% of bankroll
- • Maximum: ~2% of bankroll
- • Typical: 1u (1% of bankroll)
We cap at ~2% because of estimation uncertainty, multiple simultaneous bets, and account longevity. Most professionals avoid staking more than 1–2% per bet; 2–4% is aggressive and rarely justified for props.
The Ed Thorp Insight
Legendary mathematician Ed Thorp found that overbetting is worse than underbetting, half Kelly offers insurance against estimation errors, and the cost of reducing from full to half Kelly is only 25% growth. His advice: "When in doubt, bet less."
Practical Implementation
Step-by-Step: Using Kelly for Your Bets
- Assess True Probability — Analyze data, estimate win probability, err conservative
- Calculate Implied Probability — Implied probability = 1 / odds
- Determine Your Edge — Edge = your probability - implied probability
- Apply Kelly Formula — f* = (bp - q) / b
- Apply Fractional Kelly — Multiply by 0.1 for props, 0.25 for tennis/match markets
- Calculate Stake in Currency — Multiply percentage by bankroll
- Track & Adjust — Record bets, recalculate bankroll regularly
Kelly Calculator
Use our free Kelly calculator at /calculator. Input bankroll, decimal odds, and your win probability estimate, then choose your fractional multiplier (one-tenth or quarter Kelly). The calculator outputs your recommended stake instantly.
Golden Rules
- Never bet more than 5% of bankroll on a single bet
- Use one-tenth Kelly for props, quarter Kelly for tennis
- Recalculate bankroll monthly
- If Kelly says don't bet, don't bet — no exceptions
- Track everything
Kelly vs Other Staking Plans
Flat Staking: Simple but suboptimal — treats all bets equally.
Percentage Staking: Better than flat, but doesn't account for edge size or odds.
Martingale: Mathematically unsound — guarantees ruin with insufficient bankroll.
Kelly Criterion: Mathematically optimal for maximizing growth, automatically adjusts to edge and odds. Best long-term approach for value bettors.
The verdict: For serious bettors with genuine edges, Kelly (fractional) is superior. For recreational bettors or those without proven edges, simple percentage staking is safer.
Conclusion
The Kelly Criterion represents the most mathematically rigorous approach to bankroll management in sports betting. When applied correctly with fractional Kelly, it maximizes long-term growth while controlling risk of ruin.
But Kelly is not a magic formula. It demands accurate probability estimation, honest edge assessment, psychological discipline, and a long-term perspective. Most critically, Kelly cannot create an edge where none exists.
Our approach at Il Margine combines data-driven selection with conservative fractional Kelly staking (one-tenth for props, quarter for tennis). This balance delivers sustainable growth without the volatility that destroys most bettors.
The mathematics are clear: if you have an edge, Kelly tells you how much to bet. If you don't have an edge, Kelly tells you not to bet at all. That honesty and mathematical discipline is what separates profitable betting from gambling.
Start conservative. Track everything. Let the mathematics guide you.