Quick Summary
The Kelly criterion is a mathematical method for sizing a bet based on your edge and the odds, with the goal of maximising long-term bankroll growth while limiting the risk of ruin. It calculates the fraction of your bankroll to stake using the formula f = (bp − q) / b, where b is the net odds, p is your estimated win probability and q is the chance of losing.
- Kelly is a sizing method, so it is only as accurate as the win probability you feed into it.
- Convert decimal, fractional or American odds to decimal and subtract 1 to get b before calculating.
- A small overestimate of win probability can make the Kelly stake several times too large.
- Half or quarter Kelly reduces variance and tolerates estimation error better than full Kelly.
- Log every bet, including closing odds, to check whether your predicted win rate matches real results.
Table of Contents
- What the Kelly Criterion Is and Why Bettors Use It
- The Kelly Formula and Its Variables
- Kelly Criterion Betting Examples Across Odds Formats
- Estimating Your Edge and Why Probability Errors Matter
- Fractional Kelly Betting for Lower Variance
- Flat Betting vs Kelly Criterion: Which Staking Method Fits
- Using a Kelly Criterion Calculator and Tracking Your Results
- Start Testing Sizing Strategies on Edge's Free Trial
- Frequently Asked Questions
What the Kelly Criterion Is and Why Bettors Use It
Last Updated: October 30, 2026
The question at the heart of value betting is: how much should you stake on a bet you believe is mispriced? The Kelly criterion is the most widely cited mathematical answer.
The Kelly criterion is a sizing method, not a prediction method.
Here is the core idea. If a bet has positive expected value, the Kelly criterion tells you the fraction of your bankroll to stake.
Bettors reach for Kelly because it is one of the few staking methods with a clear mathematical objective.
Most guides present the formula as if it were a rule to follow blindly. That is the wrong way to use it.
The Kelly Formula and Its Variables
The standard formula for a binary bet is:
f = (bp − q) / b
Where:
- f is the fraction of your bankroll to stake
- b is the net odds received on the wager (decimal odds minus 1)
- p is your estimated probability of winning
If your estimated probability is 55% at decimal odds of 2.00, then b = 1, p = 0.55, q = 0.45, and f = (0.55 − 0.45) / 1 = 0.10, or 10% of bankroll.
Two variables drive almost every result. The odds set the payout, and your probability estimate sets the edge. Get either wrong and the stake is wrong.
Converting Decimal, Fractional and American Odds
Before applying the formula, convert the price to decimal odds, then subtract 1 to get b.
- Decimal: 2.50 gives b = 1.50
- Fractional: 5/2 means 2.50 in decimal, so b = 1.50
- American: +150 means decimal 2.50, so b = 1.50. A price of −200 means decimal 1.50, so b = 0.50
Sportsbooks in different markets quote in different formats, so converting correctly before any calculation matters more than most people expect.
Kelly Criterion Betting Examples Across Odds Formats
A single worked example, carried through decimal, fractional and American odds, makes the method easier to verify and harder to misapply.
The bet: you estimate a 45% chance of winning, and the price is decimal 2.50.
- Decimal: b = 2.50 − 1 = 1.50. f = (1.50 × 0.45 − 0.55) / 1.50 = 0.083, or 8.3% of bankroll.
- Fractional: 3/2 converts to decimal 2.50, so b = 1.50 and f = 8.3%. The format changes how the price looks, not what it pays.
- American: +150 converts to decimal 2.50, so b = 1.50 and f = 8.3%.
On a $1,000 bankroll that is $83. The arithmetic is identical; only the conversion step differs.
Now a favourite at −200 American, decimal 1.50, b = 0.50, with a true win probability of 70%. f = (0.50 × 0.70 − 0.30) / 0.50 = 0.10, or 10%.
Draws, Pushes and Multi-Outcome Markets
The binary formula assumes two outcomes: win or lose. Many sports markets do not fit that shape, and applying the binary formula blindly overstates the stake.
- Draws (three-way markets): in soccer 1X2 markets there are three outcomes. The clean adaptation is to treat the bet as a binary proposition against the field: your selection wins, or it does not. Set p to your estimated win probability and q = 1 − p, then use the decimal price of your selection. This is an approximation, but it is far safer than ignoring the draw.
- Pushes (whole-number lines): in spread and totals markets a whole-number line can push, returning your stake. A common approach is to reduce the effective stake by the push probability, or to use the no-push price (half-point lines) where available. If you cannot estimate the push probability, stake a fraction of the binary result rather than the full amount.
- Multi-outcome exotics: for outrights and place markets, the binary formula does not apply directly. Either model the market as a set of binary bets or use a fractional stake and treat the result as an approximation.
A Consistent Cross-Format Check
Before staking, run this three-step check:
- Convert the price to decimal odds.
- Subtract 1 to get b.
- Confirm p + q = 1 and that q reflects every way the bet can lose, including draws and pushes.
If step 3 fails, the market is not binary and the stake needs an adaptation, not the raw formula.
Estimating Your Edge and Why Probability Errors Matter
The formula assumes you know your true win probability. You do not. You estimate it, and your estimate carries error.
Suppose you believe a team has a 55% chance to win at odds implying 50%.
This is the most common way Kelly goes wrong in practice.
Edge Published results, where available, let readers check whether estimated edges translate into actual returns over time.
Fractional Kelly Betting for Lower Variance
Full Kelly maximises the expected logarithm of bankroll growth under perfect information. Real bettors do not have perfect information, and full Kelly produces large swings.
- Half Kelly stakes 50% of the full Kelly amount.
- Quarter Kelly stakes 25% of the full amount. It is more conservative and tolerates probability errors better.
Fractional Kelly is a hedge against your own estimation error. If your edge is actually smaller than you calculated, a fraction of the stake limits the damage.
The trade-off is that fractional betting grows your bankroll more slowly if your estimates are accurate.
Flat Betting vs Kelly Criterion: Which Staking Method Fits
Flat betting means staking the same unit size on every bet, regardless of edge.
Kelly does the opposite. It scales stakes to your edge, so larger edges get larger stakes. The advantage is theoretical efficiency.
Flat betting fits bettors who are still developing their models, who cannot reliably estimate probabilities, or who want predictable risk.
Many serious bettors use a hybrid.
Using a Kelly Criterion Calculator and Tracking Your Results
A calculator removes the arithmetic, not the judgment. Enter the odds and your probability, and it returns a stake. The output is only as good as the estimate you feed it, so treat the calculator as a check on your inputs rather than a source of truth.
Backtesting and Tracking Performance
Tracking is what turns a formula into evidence. A practical record for each bet should include:
- the date and event
- the market and selection
- the odds taken (in decimal) and the closing odds
Over a large sample, compare your predicted win rate with your actual win rate. If you consistently win less often than you estimated, your probabilities are inflated and your Kelly stakes are too large. Comparing the odds you took with the closing odds is a second check: if you consistently beat the close, your estimates are likely adding information; if you consistently take worse prices, the edge may be illusory.
Real-World Constraints: Limits, Odds Movement and Availability
The formula assumes you can stake the recommended amount at the quoted price. In practice:
- Sportsbook limits may cap the stake below the Kelly recommendation, especially on niche markets.
- Odds movement can mean the price is gone by the time you bet; recalculate with the available price rather than the one you modelled.
- Bet availability varies by jurisdiction and account status, so a theoretical stake may not be placeable at all.
When the available stake is smaller than the Kelly recommendation, the practical response is to take the smaller stake, not to chase the price elsewhere at worse odds.
Responsible Gambling and Loss-of-Bankroll Risk
Kelly maximises modelled long-run growth under correct inputs. It does not guarantee profit, and it does not make a bet safe. With full Kelly, drawdowns of 50% or more are a normal part of the path even when the edge is real.
Safeguards that practitioners commonly use:
- Stake caps: set a maximum percentage of bankroll per bet, regardless of what the formula returns.
- Stop rules: define a drawdown level at which you pause and review your probability estimates.
- Fractional Kelly: use half or quarter Kelly to tolerate estimation error.
Keep a bankroll record alongside the bet log. Seeing your drawdown history helps you avoid chasing losses by raising stakes, which is the fastest route to ruin.
Start Testing Sizing Strategies on Edge's Free Trial
Reading about Kelly is the easy part. Testing it on real picks with transparent results is what separates theory from evidence.
Try it with a small bankroll first, track your stakes against the formula, and compare fractional Kelly with flat betting over a meaningful sample. Sign up at Sport Betting Edge to see how sizing strategies perform on live picks before you commit real capital.
Frequently Asked Questions
What is the Kelly criterion in sports betting?
The Kelly criterion is a formula that calculates what fraction of your bankroll to stake on a bet, based on the odds offered and your estimated probability of winning. It aims to maximise long-term bankroll growth by staking more when your edge is larger and nothing when there is no edge. The result depends entirely on how accurate your probability estimate is, so it works best when paired with a disciplined process for estimating edge and tracking results over many bets.
How do you calculate a Kelly bet size?
Use the formula f* = (bp - q) / b, where b is net decimal odds minus one, p is your estimated probability of winning, and q is 1 minus p. Multiply the result by your bankroll to get the stake in currency. For example, at decimal odds of 2.00 (b = 1) with an estimated 55% win probability, f* = (0.55 - 0.45) / 1 = 0.10, so you would stake 10% of your bankroll. Check your own inputs carefully, because small errors in p change the stake substantially.
How do you use the Kelly formula with decimal odds?
Subtract one from the decimal odds to get b, the net return per unit staked. At decimal odds of 2.50, b is 1.50. Then apply f* = (bp - q) / b using your estimated probability. If the result is zero or negative, the bet has no positive expected value and the formula says to pass. Convert fractional or American odds to decimal first so the inputs are consistent.
What is fractional Kelly in sports betting?
Fractional Kelly means staking a fixed portion of the full Kelly stake, commonly one half or one quarter. Full Kelly assumes your probability estimate is exact, which it almost never is. Betting a fraction of the full stake reduces volatility and the risk of large drawdowns, at the cost of slower bankroll growth when your edge is real. Many bettors who use Kelly choose a fraction to cover the uncertainty in their own models.
Is the Kelly criterion a good strategy for sports betting?
The Kelly criterion is a sound way to size bets when you have a genuine, measurable edge and a reliable probability estimate. It is not a way to create an edge. If your model overestimates win probabilities, Kelly will tell you to bet too much, which can lead to heavy losses. Its value depends on the quality of your inputs, the discipline to track results, and practical constraints such as sportsbook limits and odds movement.
What happens if your estimated win probability is wrong?
Kelly sizing is sensitive to probability error. If you overestimate your true win rate, the formula recommends stakes larger than your real edge justifies, increasing the risk of ruin. Underestimating edge leads to smaller stakes and slower growth. This is why many bettors use fractional Kelly and why backtesting and tracking actual results over a large sample matter before committing significant capital.
How does the Kelly criterion compare with flat betting?
Flat betting stakes the same amount on every bet regardless of edge, which is simple and limits the impact of any single error in your model. Kelly stakes vary with the size of your edge and bankroll, which can produce faster growth when estimates are accurate but more volatility when they are not. Flat betting suits bettors who cannot yet estimate probabilities reliably, while Kelly suits those with validated, measurable edges.
Can you use a Kelly criterion calculator for sports betting?
Yes. A Kelly criterion calculator takes your bankroll, decimal odds and estimated win probability and returns a recommended stake, often with fractional options. Calculators save time and reduce arithmetic errors, but they do not validate your inputs. Always double-check that your probability estimate is realistic and that the odds you enter are the odds you can actually get at the sportsbook.
