Kelly Criterion in Sports Betting: Optimal Stake Sizing for Professional Bettors
Summary
- Kelly tells you the exact fraction of your bankroll to bet to maximise long-run growth
- Formula: f* = (bp − q) / b where b = net odds, p = win probability, q = loss probability
- Full Kelly is mathematically optimal but practically dangerous — most professionals use ¼ to ½ Kelly
- Kelly assumes your probability estimates are perfectly calibrated — overestimate your edge and Kelly over-bets, risking ruin
- With a 5% edge on a 1.90 bet, Full Kelly recommends ~5.3% of bankroll — ¼ Kelly = ~1.3%
The Kelly Criterion, developed by physicist John L. Kelly Jr. in 1956, solves a specific problem: given a +EV bet, how much of your bankroll should you stake? Bet too little and you under-grow. Bet too much and variance risks wiping you out.
Kelly's formula finds the mathematically optimal fraction. Used correctly — especially in fractional form — it is the foundation of professional bankroll management.
The Kelly Formula Explained
The Kelly formula determines what fraction of your bankroll (f*) to stake on a bet:
f* = (bp − q) / b
Where:
- b = net odds (decimal odds minus 1; e.g. odds of 2.10 → b = 1.10)
- p = estimated probability of winning (as a decimal)
- q = probability of losing = 1 − p
The formula can also be written as: f* = (Probability × Odds − 1) / (Odds − 1), which makes it clear that Kelly is directly derived from your EV divided by the net odds — i.e., your edge divided by the variance.
Kelly calculation: odds 2.10, estimated win probability 55%
b = 2.10 − 1 = 1.10 | p = 0.55 | q = 0.45
f* = (1.10 × 0.55 − 0.45) / 1.10 = (0.605 − 0.45) / 1.10 = 0.155 / 1.10 = 0.1409 = 14.1% of bankroll
On a €10,000 bankroll, Kelly recommends staking €1,409 on this bet. That feels large — and it is. This is Full Kelly, mathematically optimal only if your 55% probability estimate is exactly correct.
Verification: the implied probability at 2.10 odds is 47.6%. Your edge over the implied probability is 7.4 percentage points. Kelly translates this edge into a 14.1% bankroll allocation — roughly double the edge percentage, reflecting the odds structure.
Full Kelly vs Fractional Kelly
Full Kelly is the theoretically optimal fraction, but it has severe practical problems. Kelly's derivation assumes your probability estimates are perfectly accurate. In reality, even sophisticated models have estimation errors. If your true win probability is 52% and you estimate 55%, Full Kelly tells you to bet 14.1% when the correct Kelly fraction would be approximately 4.3% — a factor of 3 too high.
Overestimating your edge with Full Kelly leads to overbetting, which increases variance and can cause large drawdowns even when you have a genuine edge. Kelly itself shows that betting above the optimal fraction reduces long-run growth rate — aggressively so. Betting double the Kelly fraction produces the same long-run growth as betting zero.
Professional bettors typically use fractional Kelly — a fixed percentage of the Full Kelly recommendation — to manage this estimation uncertainty:
| Kelly Fraction | Stake (14.1% Full Kelly example) | Long-Run Growth Rate | Maximum Expected Drawdown | Typical Use Case |
|---|---|---|---|---|
| Full Kelly (100%) | 14.1% of bankroll | Optimal (theoretical) | 50%+ drawdowns common | Theory only — not recommended in practice |
| Half Kelly (50%) | 7.05% of bankroll | ~75% of Full Kelly growth | 20–30% drawdowns | Experienced bettors with calibrated models |
| Quarter Kelly (25%) | 3.5% of bankroll | ~55% of Full Kelly growth | 10–15% drawdowns | Standard professional practice |
| Eighth Kelly (12.5%) | 1.75% of bankroll | ~35% of Full Kelly growth | 5–8% drawdowns | New bettors / high uncertainty models |
Quarter Kelly is the most widely cited professional standard. It captures over half of Full Kelly's growth rate while reducing drawdowns to manageable levels — and it provides a natural buffer against probability estimation errors of up to ~3 percentage points before the strategy becomes aggressive.
Quarter Kelly applied to an AH betting portfolio over 100 bets
Setup: €10,000 starting bankroll, 5% average EV edge, average odds 1.90, estimated win probability 52.6%, Full Kelly = 5.3% per bet, Quarter Kelly = 1.33% per bet.
- Quarter Kelly stake on first bet: €10,000 × 1.33% = €133
- Expected bankroll after 100 bets at 5% EV: €10,000 × (1 + 0.05)^... — the compounding effect produces roughly €10,000 × 1.58 = ~€15,800, assuming stakes are recalculated from current bankroll each bet
- Worst-case drawdown probability (Quarter Kelly): A sequence of 15 consecutive losses — plausible at 47.4% loss probability — would reduce the bankroll by approximately 18%. The same sequence on Full Kelly would reduce it by ~55%
- After 100 bets: With genuine 5% EV and Quarter Kelly, a 52-win / 48-loss split on 1.90 bets generates approximately €2,300 profit — a 23% ROI on starting bankroll
The key takeaway: Quarter Kelly sacrifices some theoretical growth in exchange for the variance reduction that allows a bettor to survive the inevitable losing runs without panicking or going broke.
Kelly's Assumptions and Their Limits
Kelly's formula is built on several assumptions that are imperfectly met in real betting. Understanding these limits is essential to using it correctly.
Probability Estimation Error
Kelly requires accurate probability estimates. A 2-percentage-point overestimation of your win probability can double or triple the recommended stake. Since sports betting models are never perfectly calibrated, Full Kelly systematically over-bets. Fractional Kelly is the correction — it behaves as if your edge estimate is intentionally conservative.
No Simultaneous Bets Assumption
Classic Kelly assumes sequential bets with bankroll recalculation between each. In practice, professional bettors often have multiple open bets simultaneously. The Kelly formula does not account for correlated simultaneous exposures. If two open bets are on the same league and weather is a shared factor, their outcomes are correlated — and Kelly would need to be applied to the joint position, not each bet individually. In practice, most professionals simply size each bet conservatively and track total exposure as a percentage of bankroll.
Log Utility Assumption
Kelly maximises the expected logarithm of wealth — a utility function that penalises ruin exponentially. This is theoretically sensible but not everyone's actual risk preference. A bettor who genuinely prefers higher variance in exchange for faster growth might rationally use a fraction above Kelly. A bettor with limited bankroll who cannot absorb drawdowns might use a much smaller fraction. Kelly's fraction is a starting point, not a prescription.
The Solution: Fractional Kelly
All three issues above are addressed in the same way: use a fraction of Full Kelly. Quarter Kelly specifically reduces the impact of estimation errors by 75%, handles simultaneous bet exposure more safely, and produces a variance profile that almost all professional bettors find more psychologically manageable. The reduction in theoretical growth rate (from optimal to ~55% of optimal) is the cost — and for most bettors, it is the right trade-off.
Kelly in Practice: Implementation for Serious Bettors
Implementing Kelly requires three operational components: a calibrated probability model, a record-keeping system for current bankroll, and disciplined stake recalculation.
Calibration: Track your bets against closing line prices. If your model consistently finds CLV (closing line value), your probability estimates are beating the market — this is the validation you need before trusting your Kelly inputs. Without this validation, use a smaller fraction (eighth or quarter Kelly) until the sample justifies confidence.
Bankroll tracking: Kelly staking requires knowing your current bankroll precisely. Stale bankroll figures lead to incorrect stake sizes. Update after every settled bet. If multiple bets are placed daily, many professionals use a daily opening bankroll figure to avoid constant recalculation mid-session.
Minimum stake floors: Kelly can recommend very small stakes (below 0.5% of bankroll) on low-edge bets. Most professionals set a minimum stake of 0.5% to avoid booking fees and administrative friction overwhelming the theoretical bet. Kelly fractions below this floor are simply not taken — the theoretical loss from skipping these bets is negligible compared to the cost of placing many very small bets.
Accessed through a betting broker with sharp book pricing, a properly implemented fractional Kelly strategy on Asian handicap markets represents the closest thing to a mathematically rigorous bankroll management framework available to sports bettors.
The Kelly Formula
What if Kelly recommends 0% or negative?
If the Kelly formula returns 0% or negative, the bet is not +EV by your probability estimate. Don't place it. Negative Kelly literally tells you to lay the bet (bet against it), which requires a betting exchange.
Should I use fixed percentage or Kelly?
Fixed percentage (e.g., always 2% of bankroll) is simpler to implement and avoids the risk of overestimating your edge. Kelly optimises growth if your estimates are accurate. Most professionals start with fixed percentage and transition to fractional Kelly once their model is calibrated over 1,000+ bets.
Can Kelly be used for accumulator bets?
Technically yes, but accumulator bets involve correlated outcomes and compounded probability uncertainty. Most serious bettors avoid accumulators entirely — they represent negative EV in almost all cases due to bookmaker margin applied per leg.
What's the relationship between Kelly and CLV?
Kelly determines how much to bet; CLV helps verify whether your model is generating genuine edges. High CLV over large samples confirms your probability estimates are beating the market — which is the input Kelly needs.