Educational use only. This calculator is for educational purposes. It models expected values but cannot predict outcomes. Gambling involves risk and modelled results do not guarantee any return.
Kelly Criterion Stake Calculator
Enter decimal odds, your estimate of true probability, and your bankroll to compute the Kelly-criterion-recommended stake.
How to Use This Calculator
Decimal odds are the standard AU bookmaker format. Odds of 3.00 mean $3.00 returned per $1 staked (net profit $2.00).
True probability is your independent estimate of the selection’s win probability as a percentage. The bookmaker’s implied probability is 1 / decimal odds. At 3.00, the implied probability is 33.3%. If your true probability estimate is 40%, there is positive edge, and Kelly will recommend a non-zero stake.
Bankroll is the total capital you have allocated to betting. Kelly sizing is proportional: the recommended stake is a fraction of the bankroll, not a fixed dollar amount. Enter the full bankroll, not a per-bet reserve.
Worked Example
Suppose odds are $3.00, true probability estimate is 40%, and bankroll is $1,000.
- b = 3.00 - 1 = 2.00 (net return on a win)
- p = 0.40 (true probability)
- q = 1 - 0.40 = 0.60
- Full Kelly fraction f = (b x p - q) / b = (2.00 x 0.40 - 0.60) / 2.00 = (0.80 - 0.60) / 2.00 = 0.10
- Full Kelly stake = 0.10 x $1,000 = $100
- 1/4 Kelly stake = $100 / 4 = $25
The calculator shows 1/4 Kelly as the primary recommendation because in practice, true probability estimates contain error. The Kelly formula is optimal only when the probability estimate is exact. In real betting, overestimating the edge leads to overbetting and excess drawdown. 1/4 Kelly absorbs a significant amount of estimation error while still growing the bankroll proportionally on positive-EV bets.
Why 1/4 Kelly and Not Full Kelly
Full Kelly is mathematically optimal when you know the true probability exactly. In practice, you never do.
Estimation error in the true probability input creates a convex relationship with drawdown: a small error in probability produces a large error in stake size and, if the edge is overestimated, significant bankroll erosion. A standard result in the Kelly literature is that betting at full Kelly on an overestimated edge can be worse than not betting at all in terms of long-run drawdown.
1/4 Kelly (sometimes called “fractional Kelly at 0.25”) cuts the optimal stake by 75%. In simulations across realistic estimation error bands, 1/4 Kelly produces long-run growth rates close to those of full Kelly while substantially reducing maximum drawdown and the time to recovery from a losing streak. Most practitioners who use Kelly sizing in professional contexts use fractions between 0.2 and 0.5. 1/4 Kelly is a conservative and defensible default.
For a full explanation of the Kelly criterion formula and its limitations, see the Kelly criterion explained simply guide.
When Kelly Recommends Zero
If your true probability estimate implies no edge (or negative edge), the Kelly formula returns a fraction at or below zero. The calculator floors this at zero. A zero Kelly recommendation means: do not stake. The formula is explicitly telling you the bet has no positive expectation at the given probability estimate.
If you believe in the bet despite a zero Kelly output, check your probability estimate. Either the odds are worse than you thought, or the probability is lower than you thought. The formula is consistent; the estimate is the variable.
Related Reading
- Kelly criterion explained simply: the formula, history, and when to use it
- EV calculator: compute the EV of a bet before sizing it
- Dutching calculator: split a stake across multiple selections
- EV betting Australia guide: AU-specific EV betting context
For the true probability signal that makes Kelly sizing practical in AU racing, see EVSTREAM. For how we evaluate EV signal tools in our comparison universe, see our scoring methodology.