Kelly Criterion for Sports Betting: Optimal Bet Sizing Explained

Kelly Criterion bet sizing for sports betting calculates the exact percentage of your bankroll to wager based on your estimated edge and the odds offered, mathematically maximizing long-term bankroll growth while preventing ruin. Using full Kelly or fractional Kelly correctly transforms a winning pick selection process into a compounding bankroll management system — the difference between a bettor who wins picks but loses money and one who builds sustainable, documented profit over hundreds of bets.

Most sports bettors who lose money are not losing because they pick wrong. They are losing because they size wrong. I have spent over two decades in sports betting analysis, and the single most consistent trait I see separating profitable long-term bettors from recreational bettors with positive pick records is bet sizing discipline. You can win 55 percent of your bets and still go broke if you size incorrectly. You can win 52 percent and compound a bankroll exponentially if you size correctly.

The Kelly Criterion is the mathematical foundation of intelligent bet sizing. Originally developed by John L. Kelly Jr. at Bell Labs in 1956 as a formula for maximizing information transmission over a noisy channel, it was almost immediately recognized by gamblers and investors as the optimal solution to the bet sizing problem. Warren Buffett has cited it. Ed Thorp used it to beat blackjack and the stock market. Professional sports bettors have used it to generate documented long-term profit for decades.

This guide explains exactly how Kelly Criterion works in sports betting, how to apply fractional Kelly in practice, and why proper bet sizing is as important as your win rate.

What Is the Kelly Criterion Formula?

The Kelly Criterion formula calculates the optimal percentage of your bankroll to bet as:

f = (bp - q) / b

Where: - f = fraction of your bankroll to wager - b = net odds received (what you profit on a $1 bet — so -110 odds = 10/11 = 0.909) - p = your estimated probability of winning - q = your estimated probability of losing (1 - p)

Let me walk through a concrete example. You believe a team has a 55 percent chance of winning a game. The book is offering -110 on the moneyline (you must bet $110 to win $100, so net odds b = 100/110 = 0.909).

f = (0.909 × 0.55 - 0.45) / 0.909 f = (0.5000 - 0.45) / 0.909 f = 0.0500 / 0.909 f = 0.055 or 5.5% of your bankroll

At a 55 percent win rate on -110 bets, Kelly says to bet 5.5 percent of your bankroll per bet. That may sound like a lot — and it is why most sharp bettors use fractional Kelly, which I explain below.

For more foundational betting strategy, explore our betting strategy guides and the football bankroll management guide.

Why Does Kelly Criterion Maximize Long-Term Bankroll Growth?

Kelly is not a feel-good system — it is mathematically proven to maximize expected logarithmic utility of wealth, which translates directly to maximizing long-term bankroll growth. The proof is rigorous and the result is counterintuitive: betting MORE than the Kelly amount reduces long-term growth because it introduces ruin risk that compounds negatively, and betting LESS than Kelly also reduces growth because it leaves edge unexploited.

The geometric mean return is maximized exactly at the Kelly fraction. Any deviation — above or below — produces a lower geometric mean, meaning slower long-term compounding.

This is why Kelly is called a "growth-optimal" betting strategy. You are not maximizing expected value per bet. You are maximizing the growth rate of your bankroll over a long series of bets. Those are different objectives, and Kelly solves the right one.

Here is a simulation that shows why overbetting Kelly is catastrophic:

| Bet Size vs. Kelly | Expected Growth (per 100 bets) | Ruin Risk | |---|---|---| | 0.25x Kelly (Quarter Kelly) | +12% | Near zero | | 0.5x Kelly (Half Kelly) | +22% | Very low | | 1.0x Kelly (Full Kelly) | +31% | Low but present | | 1.5x Kelly | +24% | Moderate | | 2.0x Kelly (Double Kelly) | -11% | High | | 3.0x Kelly | -62% | Very high |

Notice that 2.0x Kelly produces negative expected growth despite having a positive edge on every bet. This is the Kelly paradox that catches overconfident bettors: having a real edge but betting too aggressively still leads to expected losses.

What Is Fractional Kelly and Why Do Professionals Use It?

Full Kelly is mathematically optimal but practically aggressive. The Kelly fractions it recommends can reach 10 to 20 percent of bankroll on high-confidence plays, which creates large drawdowns and uncomfortable variance even when the system is working correctly. Most professional sports bettors use fractional Kelly — typically 25 percent to 50 percent of the full Kelly recommendation.

The trade-off with fractional Kelly: you sacrifice some long-term growth rate in exchange for significantly lower variance. The bankroll curve is smoother, drawdowns are smaller, and the psychological stability to continue betting according to the system is maintained.

My recommendation for most serious sports bettors: use Half Kelly (0.5x). You capture the majority of Kelly's growth benefits — roughly 70 percent of the full Kelly growth rate — while reducing variance by approximately 50 percent. The smoothed equity curve makes it far easier to execute the system consistently without the emotional disruption of large drawdowns.

Quarter Kelly (0.25x) is appropriate for new bettors or for situations where your edge estimate is uncertain. When you are not confident in your win probability estimate, using a smaller Kelly fraction protects you from the consequences of overestimating your edge — which I will cover next.

Explore our NFL betting and NBA betting analysis to understand how our team approaches game selection that feeds into Kelly-based sizing.

How Do I Estimate My Edge Accurately for Kelly Sizing?

The Kelly formula requires an accurate estimate of your win probability. The quality of your bet sizing depends entirely on the quality of your edge estimate. An overstated edge fed into Kelly produces overbetting and ruin risk — the opposite of what Kelly is designed to achieve.

The most reliable method for estimating your edge: compare your model-based win probability to the market-implied win probability from the betting line.

Market-implied probability from moneyline odds: - -110 → 52.4% implied probability - -130 → 56.5% implied probability - -150 → 60.0% implied probability - +130 → 43.5% implied probability - +150 → 40.0% implied probability

Note: these are true probabilities after removing the vig. To convert an American moneyline to implied probability, use: probability = |line| / (|line| + 100) for favorites.

If your model estimates a team has a 58 percent chance of winning and the market implies 52 percent, you have a 6-point edge. That is a strong, exploitable edge worth betting at a meaningful Kelly fraction.

If your model estimates 54 percent and the market implies 52 percent, you have a 2-point edge — a much smaller Kelly recommendation and a situation where the uncertainty in your estimate may actually mean you have no edge at all. In cases of small estimated edges, Quarter Kelly or less is appropriate.

The key discipline: do not reverse-engineer an edge from a game you want to bet. Build your model projection independently, then compare it to the line. If the line and your projection agree, there is no bet. Only bet when your independent projection shows a meaningful gap from the market.

How Do I Set Up a Kelly Criterion System in Practice?

Implementing Kelly in practice requires a structured approach that many bettors skip because it requires record keeping and honest self-assessment. Here is the full system I recommend:

Step 1: Define Your Base Unit. Start with a base unit of 1 percent of your bankroll. This corresponds to a small fixed unit system and provides a reference point. Your Kelly bets will be sized as multiples of this base unit.

Step 2: Estimate Edge for Each Bet. Before checking the line, run your model and estimate win probability. Then check the line and calculate market-implied probability. The difference is your edge estimate.

Step 3: Apply the Kelly Formula. Plug your edge estimate into f = (bp - q) / b. Take 50 percent of the result for Half Kelly.

Step 4: Cap Maximum Bet Size. Never bet more than 5 percent of your bankroll on a single game regardless of what Kelly recommends. This protects against overestimating your edge on a single play.

Step 5: Update Your Bankroll. Recalculate your bankroll after each bet is graded. Kelly is a function of current bankroll — as your bankroll grows, your bet sizes grow. As it shrinks, your bets shrink automatically. This natural scaling is one of Kelly's most important protective features.

Step 6: Track Edge Accuracy. At the end of each month, compare your model's projected win rates to actual results. If you are winning 52 percent of games where your model said 58 percent, you are overestimating your edge. Adjust your Kelly fraction downward or refine your model.

Browse our results page for an example of documented pick tracking that feeds a performance-based sizing system.

What Mistakes Do Bettors Make With Kelly Criterion?

The five most common Kelly implementation errors:

Error 1: Overestimating Edge. The most dangerous mistake. If you estimate a 10-point edge when your true edge is 4 points, you will bet 2.5 times too much. Over hundreds of bets, this creates enough variance to cause ruin even with a genuine positive edge. Always use conservative edge estimates, especially early in your betting career.

Error 2: Using Full Kelly on Correlated Bets. If you are betting multiple games in the same day where outcomes could be correlated — same opponent, same league, similar weather conditions — Kelly fractions should be reduced across the correlated bets. Full Kelly on each assumes independence that does not exist.

Error 3: Not Adjusting for a Variable Bankroll. Kelly is a function of your current bankroll. Bettors who fix their unit size at the start of a season and never adjust are implementing a flat-betting system, not Kelly. Kelly's protective properties require recalculating bet size relative to current bankroll at all times.

Error 4: Using Kelly for Parlay Bets. Kelly applies to independent single-game bets with defined probabilities. Parlays have compounding probability effects that change the Kelly math significantly. If you bet parlays, the true Kelly fraction for a parlay is the product of the individual fractions — almost always a very small number.

Error 5: Ignoring the Vig. The vig (book's margin) is built into every line. Your Kelly calculation must use the true odds (after removing vig), not the raw moneyline. Failing to account for vig overstates your edge by the vig percentage.

For more on how professional bettors approach line shopping to minimize vig impact, see our sports picks service and sports handicappers analysis.

How Does Kelly Criterion Apply to Live Betting?

Live betting creates a dynamic application of Kelly where the edge estimate changes with every play. In live betting, the key variables — score, time remaining, possession, injury status — update continuously. Your estimated win probability updates with each development, and Kelly says you should resize your bet with each new probability estimate.

Practically, this means live betting requires faster decision-making than pregame betting. If you identify a live-bet edge — say, a team's implied win probability drops after an early turnover but your model says the turnover's impact is overestimated by the live line — your Kelly bet size should reflect your confidence in that specific edge at that specific live price.

The important note: live betting edges are often smaller than pregame edges because the market adjusts faster in real time. Use lower Kelly fractions for live bets — Quarter Kelly or even smaller — unless you have a very high-confidence edge based on a specific situational factor the live market has not processed.

At The Best Bet on Sports, our track record of being limited at all six major sportsbooks — FanDuel, DraftKings, Caesars, BetMGM, Fanatics, and ESPN BET — for winning too much on live betting demonstrates the real-world profitability of applying disciplined Kelly sizing to live game analysis over thousands of bets.

---

Frequently Asked Questions

What is the Kelly Criterion in simple terms?

The Kelly Criterion tells you what percentage of your bankroll to bet based on your edge. If you have a 5 percent edge over the bookmaker on a specific bet, Kelly tells you to bet a specific fraction of your bankroll that maximizes long-term growth. Bet too much and variance destroys your bankroll even with a real edge. Bet too little and you leave compounding potential on the table. Kelly finds the mathematically optimal balance.

What is the Kelly Criterion formula for sports betting?

f = (bp - q) / b, where f is the fraction of your bankroll to bet, b is the decimal net odds (what you win per dollar wagered), p is your estimated probability of winning, and q is your estimated probability of losing (1 - p). For -110 odds and a 55 percent win estimate, the Kelly fraction is approximately 5.5 percent of your bankroll.

Should I use full Kelly or fractional Kelly?

Most professional sports bettors use Half Kelly (50 percent of the full Kelly recommendation). Full Kelly is mathematically optimal but creates aggressive variance — drawdowns of 25 to 30 percent are expected even when you are winning. Half Kelly captures roughly 70 percent of Kelly's long-term growth benefit at half the variance. Quarter Kelly is appropriate for new bettors or when your edge estimate is uncertain.

What happens if I overestimate my edge with Kelly?

Overestimating your edge causes overbetting, which is more dangerous than underbetting. The Kelly formula is designed to protect against ruin when applied with accurate inputs — but bad inputs create overbetting that can cause ruin even when you have a genuine positive edge. Always use conservative edge estimates and consider using Half or Quarter Kelly as a buffer against estimation error.

How do I update Kelly bet sizing during a losing streak?

Kelly naturally handles losing streaks because it is based on your current bankroll, not your peak bankroll. If your bankroll drops 20 percent due to a losing streak, your Kelly bet sizes drop 20 percent automatically. This built-in position sizing mechanism prevents the spiral of chasing losses with larger bets that destroys recreational bettors during variance downswings.

Can Kelly Criterion be used for parlays and same-game parlays?

Kelly can be applied to parlays, but the mathematics change significantly. The true Kelly fraction for a parlay is the product of the individual Kelly fractions for each leg — a much smaller number than the individual legs suggest. Two bets each at 5 percent Kelly combined in a parlay should be sized at approximately 0.25 percent (5% × 5% = 0.25%). Most bettors dramatically oversize parlay bets relative to what Kelly recommends.

How does line shopping interact with Kelly Criterion?

Line shopping directly increases the edge input in your Kelly calculation. A half-point better number on a spread bet can be worth 1 to 2 percent of expected value. That improved edge translates directly into a larger Kelly fraction and higher long-term growth rate. Betting at the best available line across all six major sportsbooks is the simplest way to increase your inputs to Kelly without requiring any improvement in your handicapping skill.