Half Kelly Betting Strategy: Calculator Formula and Examples
Half Kelly takes the full Kelly stake from your edge estimate and cuts it in half, creating a more survivable staking rule for sports betting markets where probabilities are noisy.
Last updated: August 20, 2026
Key Takeaways
- Half Kelly is a fractional Kelly strategy. It uses one-half of the stake recommended by the full Kelly Criterion.
- The formula still depends on a real edge. You need the sportsbook price and your own fair win probability before any stake makes sense.
- It reduces, but does not remove, volatility. Half Kelly can still recommend large bets when the estimated edge is large.
- Risk caps remain mandatory. Your unit-size plan and maximum single-bet exposure can override the calculator.
- Track the process after every bet. Half Kelly works best when paired with closing-line and result review in a bet tracker.
What Is Half Kelly?
Half Kelly is a staking method that starts with the full Kelly formula and then applies a 50% multiplier. It is one of the most common ways bettors adapt a growth-maximizing formula to an uncertain real-world betting market.
The original Kelly framework was published by John L. Kelly Jr. in 1956 as a method for maximizing long-run logarithmic growth when the bettor has an informational advantage. That theoretical setup assumes the probability input is known well enough to trust. Sports betting rarely gives you that luxury.
Half Kelly is the compromise: it still increases bet size when your edge estimate is strong, but it lowers the damage when the model, market read, injury assumption, or prop projection is wrong.
Half Kelly Formula
Start with the full Kelly formula:
Full Kelly fraction = (bp - q) / b
b = net decimal odds, p = estimated win probability, q = loss probability
Then cut the result in half:
Half Kelly stake = Full Kelly fraction x 0.5
If full Kelly is zero or negative, half Kelly is also zero. A staking method should never be used to justify a negative-EV bet.
Example: -110 Bet With a 55% Fair Probability
Suppose your research estimates a bet should win 55% of the time, and the sportsbook offers -110. First convert -110 to decimal odds with the odds converter: decimal odds are about 1.909, so net decimal odds are 0.909.
| Input | Value | Role in the calculation |
|---|---|---|
| Bankroll | $2,000 | Stake is a percentage of this bankroll. |
| Sportsbook odds | -110 | Decimal odds 1.909; net odds 0.909. |
| Estimated win probability | 55% | Your researched fair probability. |
| Loss probability | 45% | 1 - 0.55. |
Full Kelly: (0.909 x 0.55 - 0.45) / 0.909 = 5.5% of bankroll
Half Kelly: 5.5% x 0.5 = 2.75% of bankroll
On a $2,000 bankroll, full Kelly would suggest about $110. Half Kelly reduces the stake to about $55. If your rules cap one bet at 2% of bankroll, the final stake is $40 because the risk cap is stricter than half Kelly.
Half Kelly vs Full Kelly vs Quarter Kelly
Fractional Kelly is not a different edge model. It is a different risk setting applied to the same edge estimate.
| Method | Multiplier | If full Kelly is 6% | Best fit |
|---|---|---|---|
| Full Kelly | 100% | 6.0% | Rare cases with strong calibration, high bankroll tolerance, and low model uncertainty. |
| Half Kelly | 50% | 3.0% | Tracked bettors who want meaningful edge scaling without full Kelly volatility. |
| Quarter Kelly | 25% | 1.5% | Conservative staking, early model testing, and softer manual probability estimates. |
| Flat betting | Fixed unit | Usually unchanged | Simple bankroll discipline when probability estimates are too uncertain for variable sizing. |
Half Kelly is the middle ground. It is less fragile than full Kelly, but it can still become too large for props, same-game parlays, thin edges, and correlated bets.
When Half Kelly Makes Sense
- You have a tracked probability process. Your fair prices are not guesses; they come from a model, market comparison, or repeatable research workflow.
- You review closing line value. If your bets regularly beat the close, a larger fractional Kelly setting is easier to defend.
- Your edge estimate is not tiny. Small edges can vanish after vig, line movement, limits, and model error.
- Your bankroll can tolerate swings. Half Kelly still produces drawdowns, especially when several bets are correlated.
- You can pass on weak prices. If the line moves, rerun the math instead of using the old stake.
When Half Kelly Is Too Aggressive
Half Kelly is often too large when the probability estimate is borrowed, subjective, or unstable. Player props, injury-sensitive markets, low-limit books, longshot futures, and correlated parlay legs all create ways for a clean-looking number to overstate the real edge.
Also be careful with overlapping exposure. Five half-Kelly bets tied to the same quarterback, pitcher, injury report, or weather assumption are not five independent risks. Cap exposure by game, market, and thesis.
A Half Kelly Workflow for Sports Betting
- Define the bankroll you are actually willing to risk.
- Convert the sportsbook odds with the odds converter.
- Estimate fair probability from your model, market comparison, or research process.
- Check the expected value with the EV calculator.
- Run the full stake through the Kelly calculator.
- Multiply the full Kelly output by 0.5.
- Apply the smaller of half Kelly, your unit-size cap, and your correlated-exposure cap.
- Log odds, fair probability, stake, result, closing line, and notes in a bet tracker.
How to Adjust Half Kelly for Model Error
The cleanest way to use half Kelly is to discount the probability input before sizing the bet. If your model says 56% at -110, do not blindly size from 56% every time. Ask how often similar historical edges were actually realized.
Three practical adjustments help:
- Probability haircut: reduce your estimated win probability before running Kelly when the input is uncertain.
- Stake cap: limit any one bet to a fixed percentage, such as 1% or 2%, even when the calculator recommends more.
- Market confirmation: compare books and closing prices with line shopping and CLV tracking.
Those adjustments make half Kelly less elegant but more useful. Sports betting rewards process survival before it rewards theoretical precision.
Size the Bet After You Audit the Edge
Juice can analyze a bet slip screenshot, compare the sportsbook price to an AI-estimated probability, and help you decide whether a half-Kelly stake is too aggressive.
Download Juice on iOSFAQ
What is half Kelly betting?
Half Kelly betting means calculating the full Kelly stake and then betting 50% of that amount. If full Kelly recommends 4% of bankroll, half Kelly recommends 2%.
How do you calculate a half Kelly stake?
Calculate the full Kelly fraction with f = (bp - q) / b, then multiply the result by 0.5. In the formula, b is net decimal odds, p is win probability, and q is loss probability.
Is half Kelly better than full Kelly?
Half Kelly is usually safer for sports bettors because full Kelly assumes the probability estimate is accurate. Half Kelly reduces volatility and leaves room for model error, stale injury assumptions, and market noise.
When should you use half Kelly instead of quarter Kelly?
Use half Kelly when your probability estimates are tracked, your bankroll is clearly defined, and your exposure cap can tolerate larger bets. Use quarter Kelly when your edge estimate is softer or you are still testing the process.
Can half Kelly still lose money?
Yes. Half Kelly only controls stake size after you estimate a positive edge. If the edge estimate is wrong, the odds are stale, or the bets are correlated, half Kelly can still produce losses and drawdowns.
Sources and References
- John L. Kelly Jr.: A New Interpretation of Information Rate
- Edward O. Thorp: Optimal Gambling Systems for Favorable Games
- American Gaming Association: Responsible Play
- National Council on Problem Gambling: Responsible Sports Gambling
BetResearcher is an independent research site. This guide is educational, not financial advice or a guarantee of profit. Sports betting involves risk, and you should only bet where legal and with money you can afford to lose.