Kelly Criterion Calculator
Enter your win rate and how much you win and lose each time (or the decimal odds) to get the Kelly fraction: the share of your capital to bet each time so that it grows fastest in the long run. The calculator also shows half and quarter Kelly, the win/loss ratio, the break-even win rate, the expected value and a chart of bet size against long-term growth.
How to use the Kelly criterion calculator
- Choose how to enter the numbers: Gain and loss % (investing) for trades where a loss doesn't wipe out the whole stake (for example take profit at +15% and stop loss at −5%), or Decimal odds (betting) for sports betting and games where you lose the whole stake and a win pays stake × odds.
- Enter your win rate and how much you win and lose (or the odds). Add your capital if you want the bet sizes as amounts.
- Press Calculate or Enter. You get the Kelly fraction, half Kelly and quarter Kelly, plus the win/loss ratio, the break-even win rate, the expected value and the average growth per bet.
- The chart plots bet size against long-term growth; its peak is the Kelly fraction. If you change any input afterwards, the result fades and reminds you to calculate again.
What is the Kelly criterion?
The Kelly criterion was published in 1956 by John Kelly of Bell Labs. It answers one question: when you know the probability of winning and the payoff, and you can bet again and again, what share of your capital should you bet each time so that it grows fastest over the long run?
Bet too little and your money grows slowly. Bet too much and a few losses in a row set you back so far that you end up with less in the long run. The Kelly criterion finds the fraction in between that maximizes the long-term compound growth rate.
Kelly formula for bets (lose the whole stake)
- f*: fraction of capital to bet
- p: probability of winning; q: probability of losing (q = 1 − p)
- b: net odds, what you win per 1 staked (decimal odds − 1)
Example: win rate 60% at even money (b = 1): (1 × 0.6 − 0.4) ÷ 1 = 20% of your capital per bet.
Kelly formula for investing (partial losses)
- a: loss when you lose (your stop loss)
- b: gain when you win (your take profit)
With a = 100% this is the same as the betting formula.
Example: win rate 40%, take profit +15%, stop loss −5%: 0.4 ÷ 0.05 − 0.6 ÷ 0.15 = 8 − 4 = 400%. That looks alarming, but it is a position size: each loss is only 5%, so one losing trade costs 400% × 5% = 20% of your capital. A position above 100% needs leverage, which hardly anyone uses in practice, and that is one more reason to bet half Kelly.
Win rate and win/loss ratio table
Win/loss ratio = average gain ÷ average loss. The table shows the Kelly fraction when a loss costs the whole stake (a = 100%); a dash means the expected value is not positive and you should not bet:
| Win/loss ratio | Break-even win rate | Win rate 40% | 45% | 50% | 55% | 60% |
|---|---|---|---|---|---|---|
| 1 : 1 | 50% | — | — | — | 10% | 20% |
| 1.5 : 1 | 40% | — | 8.33% | 16.67% | 25% | 33.33% |
| 2 : 1 | 33.33% | 10% | 17.5% | 25% | 32.5% | 40% |
| 3 : 1 | 25% | 20% | 26.67% | 33.33% | 40% | 46.67% |
Break-even win rate = 1 ÷ (1 + win/loss ratio). At 2 : 1 you only need to win more than a third of the time to have a positive expected value; at 1 : 1 you need more than half.
Negative expected value: don't bet
If the Kelly criterion gives 0 or a negative number, the bet does not have a positive expected value. Whatever fraction you choose, you lose money in the long run, so the best move is not to bet.
For example, a sports bet at odds of 1.90 that you think is a coin flip: (0.9 × 0.5 − 0.5) ÷ 0.9 ≈ −5.56%. On average you lose 5 for every 100 staked, which is the bookmaker's margin. Kelly betting only makes sense if your estimated win rate is genuinely higher than the probability implied by the odds.
Why most people bet half Kelly
The Kelly fraction gives the fastest long-term growth, but it is not a comfortable ride:
- A wrong win rate costs a lot: nobody knows the true win rate. Overestimate it and the fraction comes out too large. Betting more than Kelly only adds risk without adding return, and at about twice the Kelly fraction long-term growth falls to zero; beyond that you lose money in the long run (you can see this on the calculator's chart).
- Full Kelly swings hard: by a standard theoretical approximation, betting full Kelly gives about a 50% chance that your capital halves at some point; at half Kelly the chance drops to about 12.5%.
- Half Kelly gives up little: half Kelly keeps about 75% of the long-term growth rate of full Kelly. With a 60% win rate at even money, full Kelly grows 2.03% per bet on average and half Kelly 1.52%.
That is why many people bet half Kelly or quarter Kelly and treat the Kelly fraction as a ceiling not to exceed.
The Kelly criterion for stocks and ETFs
Stock returns are not two outcomes of "win this much, lose that much" but continuous ups and downs. For them the continuous version of the Kelly criterion is used:
- μ: expected annual return; r: risk-free rate (for example a savings account or Treasury bills)
- σ: annual volatility (standard deviation of returns)
Example: expected return 8%, risk-free rate 2%, volatility 20%: (0.08 − 0.02) ÷ 0.2² = 150%, meaning 1.5× leverage in theory; half Kelly is 75%, with the rest in cash or bonds.
The result is very sensitive to the expected return: lower it from 8% to 6% and the fraction falls from 150% to 100%. For stocks the Kelly criterion works best as a guide to the maximum position size, not as a precise answer.
Limitations of the Kelly criterion
- You need the win rate and payoffs: casino games have fixed odds, but for investments they can only be estimated from past records or backtests, and the future may differ.
- It assumes endless repetition: Kelly maximizes long-run growth; over a small number of bets the results can still differ a lot from what you expect.
- It assumes independent bets: if your trades are correlated (for example several stocks in the same industry), the real risk is larger than the formula assumes.
- It ignores costs: commissions, slippage and taxes reduce your real win/loss ratio, so subtract them first.
Frequently asked questions
How do you calculate the Kelly criterion?
When a loss costs the whole stake, Kelly fraction = (net odds × win rate − loss rate) ÷ net odds. For a 60% win rate at even money, (1 × 0.6 − 0.4) ÷ 1 = 20% of your capital per bet. When a loss costs only part of the stake, use win rate ÷ loss size − loss rate ÷ gain size.
What does a negative Kelly fraction mean?
It means the expected value is not positive, so betting loses money in the long run and the Kelly criterion says not to bet. For example, odds of 1.90 with a 50% win rate give −5.56%.
What does a Kelly fraction above 100% mean?
It means that in theory you should invest more than your whole capital, which requires leverage. It is common for trades with a tight stop loss: a 5% stop loss, 15% take profit and 40% win rate give 400%. A slightly wrong win rate is magnified by leverage, so in practice use half Kelly or less and stay within the leverage you can handle.
How do you calculate the win/loss ratio?
Win/loss ratio = average gain per win ÷ average loss per loss. With a 15% take profit and a 5% stop loss it is 3 : 1. The break-even win rate is 1 ÷ (1 + win/loss ratio), so at 3 : 1 any win rate above 25% has a positive expected value.
Can I use the Kelly criterion for sports betting?
Yes. Choose decimal odds and enter the odds and your estimated win rate. Bookmaker odds already include a margin, so your win rate has to beat the probability implied by the odds (above about 52.63% at odds of 1.90) to get a positive result. If your win rate is just a gut feeling, the bet size will not be reliable either.
What is half Kelly and why is it popular?
Half Kelly means betting half the Kelly fraction. It keeps about 75% of the long-term growth rate of full Kelly with much smaller swings: in theory full Kelly has about a 50% chance of halving your capital at some point, half Kelly about 12.5%. Because the win rate is hard to estimate, most people use half Kelly or less.
Is the Kelly criterion suitable for stocks?
It is useful as a guide to the maximum position size. For stocks use the continuous version, (expected return − risk-free rate) ÷ volatility², but the expected return is hard to estimate and the result is very sensitive to it, so do not bet heavily on the exact number.
- Compound Interest Calculator: see how a growth rate compounds over the years.
- CAGR Calculator: turn a total return into an annualized return.
- IRR Calculator: annualized return when money goes in and out more than once.