Set a limit before choosing a stake
Start with what you can afford to lose, after bills, repayments and everyday spending. That amount may be zero. A staking system cannot answer the affordability question for you.
If you choose to bet, set a money limit and a time limit before the session. Losing does not increase what you can afford. Adding money to get back to even moves the limit just when it is meant to hold.
A useful record includes the amount deposited, stakes accepted, settled returns and withdrawals. Return and profit are different: a winning return includes money you originally staked.
What is a betting unit?
A unit is a reference amount used to describe stake sizes. It is not a universal amount or a recommendation.
For illustration only, a hypothetical $1,000 bankroll divided into 100 units makes one unit $10. A $20 stake is two units and 2% of that starting bankroll. If the bankroll falls to $800, the same $20 stake is 2.5% of the remaining amount.
Neither 1% nor 2% is automatically an appropriate stake. The affordable limit, total exposure and uncertainty all matter.
Check what a return actually means
The odds calculator separates the original stake, potential profit and total return. Changing an example stake does not recommend that amount.
What is variance?
Variance describes how results can fluctuate around an expected average. Even with an accurate positive-value estimate, losses can cluster. Without an accurate estimate, a supposed edge may not exist at all.
Several open bets can also be related. Two bets linked to the same player, game or team may not provide independent exposure. Looking only at each stake can understate the total risk.
A winning run does not validate every assumption. A losing run does not make the next bet more likely to win.
What does the Kelly criterion do?
For a simple binary bet without a push, one version of the formula is (p × d − 1) ÷ (d − 1), where p is the estimated win probability and d is decimal odds.
Edward O. Thorp describes the binary formula and its assumptions in The Kelly Criterion in Blackjack, Sports Betting, and the Stock Market, section 2. The source was checked on 18 October 2026.
It depends critically on the probability estimate. If p is wrong, the calculated fraction can be much too large. The formula also assumes conditions that may not hold for real betting, including the way opportunities and wealth evolve. Fractional Kelly reduces the fraction used but does not repair a bad estimate or make losses impossible.
This is an explanation of a formula, not a personal staking recommendation. There is no need to use Kelly to keep a simple affordable limit.
Test the probability assumption
At hypothetical odds of 2.00, a 55% estimate produces a full Kelly fraction of 10%. Quarter Kelly makes that 2.5%, or $25 from a hypothetical $1,000 bankroll. A losing result loses that $25.
If the estimate changes to 50%, the formula produces zero. The calculator below shows this sensitivity beside the stake. It does not estimate the true chance or select a suitable amount for you.
The full Kelly calculator explains cent rounding and the excluded settlement conditions. Its range is a set of scenarios, not a confidence interval.
Keep the limit usable
Keep records that distinguish settled results from open stakes. Track withdrawals as well as deposits. Review total exposure, including correlated bets. Set limits while calm and avoid changing them to chase a result.
You do not have to keep betting to test a method or finish a record. If keeping to the limit is difficult, pause and seek support. Gambling Help Online provides Australian gambling support information.
For the underlying numbers, read betting odds explained. For why a favourable estimate is still uncertain, read expected value betting.