Kelly criterion calculator
The stake size that grows a bankroll fastest in the long run — and why most bettors should deliberately bet less than it says.
The bet and the bankroll
Percentages go in as numbers: 55 means 55%.
Recommended stake
The Kelly criterion maximises long-run growth, not comfort. Even at the correct stake, drawdowns of 30% or more are ordinary, and the formula assumes your probability estimate is accurate — if it is optimistic, Kelly will systematically overbet and the bankroll can be lost. Never stake money you need.
Kelly sizes the bet to the edge, not to your confidence — and almost everyone should deliberately bet less than it says.
Sizing a bet so the bad run does not end it
The Kelly criterion answers what expected value ignores: given an edge, how much should you actually risk? Bet too little and a real edge compounds too slowly to matter. Bet too much and the maths turns against you — a bankroll that halves needs a 100% gain to recover, so oversized stakes destroy growth even when every single bet is favourable. The optimal fraction is the edge divided by the net odds, and it scales with the edge, not with your confidence: a 10% edge at 2.00 asks for 10% of the bankroll, the same edge at 5.00 asks for only 2.5%.
Almost nobody should bet full Kelly, because the formula assumes your probability is exactly right and it never is. Feed it an overestimated edge and it will happily recommend a stake too large for the edge you actually have — the fastest way to lose a bankroll while being right on average. Half Kelly gives up a quarter of the theoretical growth and roughly halves the volatility; quarter Kelly is what many professionals with real models use. And when the price offers no edge, the correct stake is zero: a negative Kelly fraction means the bet belongs on the other side, never that you should bet less on this one.
What that means with 1,000 in the bank
- Bankroll
- 1,000
- Price · your probability
- 2.00 · 55%
- Edge
- +10%
- Full Kelly
- 10% → 100 per bet
- Half Kelly
- 5% → 50 per bet
Five losses in a row — one sequence in 54 at this price — leave 590 at full Kelly, which then needs +69% to get back to 1,000. The same run at half Kelly leaves 774, needing +29%. Same edge, same bets, half the hole.
Questions
Should I use full, half or quarter Kelly?
Half or quarter for almost everyone. Full Kelly is only optimal if your probability estimate is exactly right, and estimation error hurts far more on the upside of the stake than it helps. Fractional Kelly trades a little growth for a lot of survivability.
What is expected growth per bet?
The average logarithmic growth of the bankroll for one bet at that stake. It is small by design — a 0.005 figure means about half a percent of compounding growth per bet — and it is the quantity Kelly maximises.
What if I have several bets at the same time?
Single-bet Kelly overstakes when bets run simultaneously, especially if they are correlated. As a practical rule, divide the total across concurrent positions and treat correlated bets as one.
Why does the calculator show zero when I think I have an edge?
Because at the price you entered, your probability does not clear the break-even point. Check the price against 1 divided by your probability: if the price is shorter, there is no edge to stake.
Other free tools
Odds converter
Betting margin calculator
EV calculator
Probability calculator
Arbitrage calculator
Parlay calculator
ROI calculator
Yield calculator
Stake calculator
Bankroll calculatorThe model’s call of the day, free on Telegram
No account, no email. Every day we publish the match where our model disagrees most with the market, 15 minutes before it starts — and we settle every pick we publish, won or lost.
Open the channel →