Kelly Criterion for Sports Betting: The Optimal Bet Size Formula

Most bettors size their bets on vibes. "I'm feeling good about this one — let's bump it to 5%." "I'm down on the day, gotta press the next one." This isn't strategy. It's emotion dressed up as decision-making.

The Kelly Criterion is the actual math behind how much to bet on each wager. It's a formula derived in 1956 by Bell Labs researcher John Kelly Jr. that calculates the bet size that maximizes long-term bankroll growth — given your edge and the odds you're betting at. Professional gamblers, hedge fund managers, and anyone who manages money under uncertainty use some version of Kelly to size positions.

This guide explains what Kelly Criterion is, how to calculate it for sports betting, why almost no one should use full Kelly in practice, and how to apply fractional Kelly correctly.

What is the Kelly Criterion?

The Kelly Criterion is a formula that determines the bet size, as a percentage of your bankroll, that maximizes long-term geometric growth given a known edge.

The key phrase is "geometric growth." Most bet sizing decisions optimize for the wrong thing — they try to maximize expected profit on a single bet, which leads to over-betting. Kelly optimizes for the compound rate of return across many bets, which is what actually matters when your bankroll has to survive long enough to realize your edge.

Bet too small relative to your edge and you leave money on the table. Bet too big and you go broke before your edge plays out. Kelly is the size that balances those two failure modes mathematically.

The Kelly formula for sports betting

For a single bet at fixed odds with a binary outcome (win or lose), the Kelly formula is:

f* = (bp - q) / b

Where:

• f* = fraction of your bankroll to bet
• b = decimal odds minus 1 (i.e., your profit per $1 staked if you win)
• p = your estimated probability of winning
• q = probability of losing (1 - p)

Worked example

You bet a moneyline at +150 (decimal odds 2.50), and you estimate your true probability of winning at 45%.

b = 2.50 - 1 = 1.50
p = 0.45
q = 0.55

f* = (1.50 × 0.45 - 0.55) / 1.50
   = (0.675 - 0.55) / 1.50
   = 0.125 / 1.50
   = 0.0833
   = 8.33%

Full Kelly says you should bet 8.33% of your bankroll on this wager. On a $5,000 bankroll, that's $416.

That number probably feels uncomfortably large. You're not wrong — and that's exactly why almost nobody should use full Kelly in practice.

Why you shouldn't use full Kelly

Full Kelly is mathematically optimal only if three things are true:

1. Your edge estimate is exactly correct. You know your true win probability with perfect precision.
2. You can withstand massive variance. Full Kelly produces roughly a 33% chance of being down 50%+ at some point, even with a real edge.
3. Your edge will persist indefinitely. Markets won't move against you, your model won't decay, your books won't limit you.

In sports betting, none of these three things are true. You don't know your true probability — you're estimating it. The variance is brutal — full Kelly drawdowns can wipe out years of gains. And edges decay constantly as markets adjust.

This is why every professional bettor and academic paper on Kelly recommends fractional Kelly — using a fraction (typically 1/4 to 1/2) of the full Kelly bet size.

The math behind why: betting half-Kelly captures roughly 75% of the long-term growth rate of full Kelly, with about a quarter of the variance. You give up a small amount of expected growth in exchange for dramatically lower drawdowns and dramatically lower risk of ruin. For real-world sports bettors who can't perfectly estimate edge, half-Kelly is generally agreed to be the sweet spot.

Quarter-Kelly is even more conservative — about 56% of full Kelly's growth rate with one-sixteenth the variance. Most professional sports bettors operate somewhere in the 1/4 to 1/2 Kelly range.

Kelly with edge uncertainty

Here's a counterintuitive truth that breaks most casual Kelly applications: if you're not certain about your edge, you should bet less than full Kelly even if your edge estimate is the best one possible.

Imagine you think a bet is 55% to win at -110 odds. Your central estimate gives you a positive edge. But what if your true win rate is only 53%? Or 51%? Your Kelly stake under those slightly-worse-than-expected scenarios drops significantly — and at 50%, Kelly says don't bet at all.

When your edge estimate has uncertainty (it always does), full Kelly becomes overconfident. You're sizing as if your point estimate is the truth, when in reality your edge could be anywhere in a range. Fractional Kelly is the practical fix for this — it builds in a margin of safety against being slightly wrong about your edge.

A useful rule of thumb: the more uncertain you are about your edge, the smaller the Kelly fraction you should use. New bettor with no CLV history? Maybe 1/8 Kelly. Established bettor with hundreds of bets and confirmed positive CLV? 1/2 Kelly is reasonable.

What's a typical Kelly bet size for a real bettor?

Sharp bettors generally have edges of 1-4% at standard -110 odds. Let's see what Kelly says for that range:

| Win Rate | Edge at -110 | Full Kelly | Half Kelly | Quarter Kelly | |---|---|---|---|---| | 51.0% | -2.7% | DON'T BET | DON'T BET | DON'T BET | | 52.4% | 0% (breakeven) | DON'T BET | DON'T BET | DON'T BET | | 53.0% | 1.2% | 1.3% | 0.65% | 0.32% | | 54.0% | 3.1% | 3.3% | 1.65% | 0.83% | | 55.0% | 5.0% | 5.2% | 2.6% | 1.3% | | 57.0% | 8.8% | 9.1% | 4.55% | 2.28% |

(At -110 odds, you need to win 52.4% to break even after vig — that's where the negative edge ends.)

Most professional sports bettors operate in the 53-55% true win rate range against -110 lines. Their typical position size is 1-3% of bankroll per bet. That's the realistic answer to "how much should I bet" when the math is honest.

If you're betting 5%+ per wager, you're either drastically over-confident in your edge, you're using full Kelly when you shouldn't, or you're sizing on emotion. None of those are good.

Kelly vs. flat unit sizing

Many bettors use flat unit sizing — same dollar amount every bet, regardless of edge. There's nothing wrong with this, and it's much simpler than Kelly. But it has trade-offs.

Flat sizing pros:

• Easy to execute, no math required per bet
• Removes emotional decision-making
• Robust against edge estimation errors

Flat sizing cons:

• Doesn't scale up on high-edge bets or down on low-edge bets
• Slower long-term growth than properly-applied Kelly
• Doesn't naturally adjust to bankroll growth over time

Kelly sizing pros:

• Mathematically optimal growth (with fractional Kelly correction)
• Scales bet size with edge — big edge means big bet, small edge means small bet
• Naturally adjusts as bankroll grows or shrinks

Kelly sizing cons:

• Requires accurate edge estimates (most bettors can't reliably produce these)
• More variance, larger drawdowns
• Easy to apply incorrectly — full Kelly will blow up most bettors

For most recreational bettors, flat unit sizing at 1-2% per bet is genuinely fine. Don't let "Kelly is mathematically optimal" make you feel bad about a simpler approach. The optimal strategy you actually execute beats the optimal strategy you abandon.

When Kelly genuinely helps

Kelly becomes meaningfully better than flat sizing when:

1. You have well-validated edge estimates (hundreds of bets with confirmed positive CLV)
2. Your edge varies meaningfully across bets (some bets are +1%, others are +5%)
3. You're sized small enough that variance won't break you (fractional Kelly, 1/4 to 1/2)

If you're brand new to betting, you don't yet meet condition #1. Track CLV across hundreds of bets first. Once you have data confirming a real edge, then start sizing with Kelly.

How to use Kelly correctly

The practical steps:

1. Estimate your win probability for the specific bet (this is the hard part — most bettors are bad at this)
2. Convert your odds to decimal, then to b (decimal - 1)
3. Calculate full Kelly using f* = (bp - q) / b
4. Take a fraction — start with 1/4 Kelly until you have hundreds of bets of CLV data, then move to 1/2 Kelly if your edge proves out
5. Cap your maximum bet at 5% of bankroll regardless of what Kelly says (sanity check against overconfidence)
6. Recalculate Kelly off your current bankroll, not your starting bankroll, after wins and losses

Step 1 is where 99% of bettors get tripped up. Estimating true win probability is genuinely hard, and most casual bettors overestimate their edge by a wide margin. Your CLV data over hundreds of bets is the real check on whether your edge estimates are calibrated.

The free Kelly Criterion calculator

We built a free Kelly calculator at truelinebets.com/tools/kelly-criterion-calculator. Plug in your win probability, odds, bankroll, and chosen Kelly fraction (full, half, or quarter) — get your bet size instantly.

It's free, no signup, no rate limit. The math runs in your browser.

The calculator is the formula. The version inside TrueLine suggests Kelly bet sizes based on your actual edge as measured by your CLV history — not an edge you guess at. The difference between guessing your edge and measuring it from real data is enormous, and that measurement is the core of why TrueLine exists.

The bottom line

Kelly Criterion is the math behind how much to bet on each wager. The formula tells you the size that maximizes long-term growth. But:

• Don't use full Kelly. It assumes your edge is known with certainty (it isn't) and produces brutal variance.
• Use fractional Kelly. Quarter-Kelly to half-Kelly is the practical range for sports bettors.
• Sanity-check with a max bet cap. Even if Kelly says bet 12%, don't bet more than 5%.
• Validate your edge first. Track hundreds of bets and check your CLV before you start sizing aggressively.
• Flat unit sizing is fine for most people. Kelly is more powerful when applied correctly, but more dangerous when applied incorrectly.

The bettors who blow up using Kelly are the ones who use full Kelly with optimistic edge estimates. The bettors who win with Kelly are the ones who use fractional Kelly with edge estimates validated by data.

Get the math right. Bet the size that compounds your bankroll, not the size that thrills you.