The Kelly Criterion for Sports Betting Explained

Kelly tells you the bet size that maximizes the long-run growth rate of your bankroll given your edge and the odds. It is the closest thing sports betting has to an optimal money management formula. It also has sharp edges most bettors do not appreciate.

Here is the formula, a worked example, and the practical reasons most professionals use a fraction of what Kelly recommends.

The formula

The Kelly Criterion for a binary bet (win or lose) is:

f* = (bp - q) / b

where:

• f* is the fraction of your bankroll to wager
• b is the decimal odds minus 1 (your profit per dollar staked on a win)
• p is your estimated probability of winning
• q is 1 - p (probability of losing)

If the result is zero or negative, the bet has no edge and Kelly says do not bet.

Worked example

Say you are looking at a -110 line. Your model says the true win probability is 55%.

Convert -110 to decimal odds: -110 means risk 110 to win 100, so decimal odds are 1.909. That makes b = 0.909.

p = 0.55, q = 0.45.

f* = (0.909 x 0.55 - 0.45) / 0.909 f* = (0.500 - 0.45) / 0.909 f* = 0.050 / 0.909 f* = 0.055 = 5.5% of bankroll

On a $1,000 bankroll, full Kelly says bet $55.

If your edge were larger, say 60% true probability, the recommendation jumps:

f* = (0.909 x 0.60 - 0.40) / 0.909 f* = (0.545 - 0.40) / 0.909 f* = 0.160 = 16.0% of bankroll

That is $160 of a $1,000 bankroll on a single bet. This is where Kelly starts to feel uncomfortable.

Why full Kelly causes drawdowns

Full Kelly is the bet size that maximizes the geometric growth rate of your bankroll. It is mathematically aggressive. The standard result is that a positive-EV bettor on full Kelly will, with high probability, experience a 50% drawdown at some point in their lifetime.

A 50% drawdown means watching your $10,000 bankroll fall to $5,000 before recovering. This is normal Kelly behavior, not a sign anything is wrong. Most humans cannot tolerate it. They cut bet sizes after the drawdown, miss the recovery, and end up underperforming Kelly anyway.

The other risk with full Kelly is estimation error. Kelly assumes your probability estimates are correct. If you are off, full Kelly amplifies the mistake. A bettor who thinks they have a 5% edge but actually has a 0% edge will, on full Kelly, lose money fast.

Half-Kelly and quarter-Kelly

Practitioners almost universally bet a fraction of the Kelly recommendation. The standard fractions:

• Half-Kelly: Multiply Kelly's f* by 0.5. Captures roughly 75% of the long-term growth rate with substantially smaller drawdowns. Most professional sports bettors use this as a default.
• Quarter-Kelly: Multiply by 0.25. Captures roughly 50% of growth with minimal drawdown risk. Common for bettors with less confidence in their probability estimates.

Reverting to the 55% / -110 example: full Kelly recommends 5.5% of bankroll. Half-Kelly says 2.75%. Quarter-Kelly says 1.4%. The growth rate gives up some compounding in exchange for emotional sustainability.

There is no purely mathematical reason to prefer one fraction over another. The choice is a personal tradeoff between growth and volatility tolerance.

When flat betting wins

For bettors who do not have well-calibrated probability estimates (which is most people), flat betting at 1-2% of bankroll often outperforms any Kelly variant in practice. The reason is simple: flat betting is robust to estimation error.

If your probability estimates are noisy, Kelly will sometimes recommend large bets on what turn out to be poor edges and small bets on what turn out to be strong ones. Flat betting treats all bets the same, which loses some theoretical efficiency but avoids catastrophic mis-sizing.

A reasonable rule of thumb: use Kelly when you have a model and historical data showing your probability estimates are accurate within a few percentage points. Use flat betting when you are estimating from intuition or limited data.

Capped Kelly in practice

Even bettors with strong estimates almost always cap their bets at a fixed maximum (typically 3-5% of bankroll), regardless of what Kelly recommends. This protects against:

• Estimation error on high-edge plays (a "60% true probability" estimate that turns out to be 53%)
• Correlation across multiple bets the same day
• Sportsbook limits on individual wagers
• Plain emotional sustainability

Capped Kelly is the bet size: f* = min(Kelly recommendation, fixed cap). The cap is rarely binding on small edges, but it prevents disasters on rare high-edge plays where Kelly might recommend 15-20% of bankroll.

Putting it together

A practical sizing approach for most bettors:

1. Estimate your edge honestly. If you cannot articulate why a bet is +EV, do not size it with Kelly.
2. Apply the formula. Compute full Kelly.
3. Multiply by 0.25 to 0.5 (your chosen fraction).
4. Cap the result at 3-5% of current bankroll.
5. Recompute as your bankroll changes. Bet sizes scale with the bankroll.

This produces conservative, math-anchored sizing that survives bad runs and compounds during good ones. It is not glamorous. It works.

The daily card recommends sizing using quarter-Kelly with a 15% daily exposure cap. The track record shows what consistent sizing looks like over a long sample.

Frequently asked questions

What is the Kelly Criterion in sports betting?

Kelly Criterion is a formula that calculates the optimal fraction of your bankroll to risk on a bet, given your estimated edge and the odds offered. It maximizes long-term bankroll growth, but at the cost of high short-term volatility.

What is the Kelly formula for sports bets?

f* = (bp - q) / b, where f* is the fraction of bankroll to bet, b is the decimal odds minus 1 (your profit per dollar), p is your estimated win probability, and q is 1 - p (loss probability). If the result is negative or zero, do not bet.

Why do professionals use half-Kelly or quarter-Kelly?

Full Kelly is mathematically optimal but extremely volatile. A 50% drawdown is statistically common even for a positive-EV bettor on full Kelly. Half-Kelly retains roughly 75% of the long-term growth with much smaller drawdowns; quarter-Kelly trades growth for stability.

Is flat betting better than Kelly for beginners?

For most recreational bettors, yes. Kelly requires accurate probability estimates. If your estimates are off, full Kelly can amplify losses fast. Flat-betting 1-2% per play removes estimation error from sizing while preserving most of the bankroll-management benefit.

Can Kelly tell me to bet more than my bankroll?

Mathematically, full Kelly can recommend bet sizes above 25% of bankroll on high-edge plays. In practice, no real bet should exceed 5% of bankroll regardless of what Kelly says. Capped Kelly (the lesser of Kelly and a fixed cap) is the standard professional approach.