Minimalist Trading
  1. Home
  2. Tools
  3. Kelly Criterion Calculator

Kelly Criterion Calculator for Trading and Betting

Find the risk per trade that maximises long-run growth, and what half or quarter Kelly costs you.

Full Kelly risk

–

Your stake or risk

–

In money

–

Growth per trade

–

Of maximum growth

–

Growth versus risk per trade

Growth rises to a peak at full Kelly, then falls. Past about twice Kelly (the red zone) you lose money even with a winning strategy.

What each fraction costs you

FractionRisk per tradeGrowth, 100 tradesTypical max drawdownBad luck (1 in 20)

Drawdowns from a 200-trade simulation of your numbers.

This Kelly formula calculator turns a win probability and a payoff, or a win probability and bookmaker odds, into the stake that grows a bankroll fastest. Compare full, half and quarter Kelly, see the stake in money, and see what each costs in drawdown.

Quick answer. Kelly stake = (b × p − q) ÷ b, where p is your win probability, q = 1 − p and b is the net payoff per unit risked (decimal odds − 1 for a bet). For a 55% win probability at decimal odds 2.10 that is 14.1% of the bankroll, so most people stake half or a quarter of it.

Understand the result. Kelly criterion for trading

What the Kelly criterion is

The Kelly criterion answers one question: given an edge, what fraction of my capital should I risk on each trade to maximise long-term growth? Bet too little and you grow slowly. Bet too much and volatility eats your growth, until past a certain point you lose money even though the strategy is profitable.

The growth curve in the calculator makes this visible. It rises to a peak at full Kelly and falls after it. At roughly twice Kelly, long-run growth drops to zero, and beyond that it is negative.

Why traders use half or quarter Kelly

  • Your inputs are estimates. If your true win rate is lower than you think, full Kelly is already over-betting.
  • Drawdowns are large. Full Kelly produces deep, frequent drawdowns. The table shows typical and bad-luck drawdowns for each fraction from a simulation of your numbers.
  • You keep most of the growth. Half Kelly delivers roughly three quarters of full Kelly's growth with far smaller swings.

Reading the result

The risk per trade shown is the share of your account you lose if a trade hits its stop (with the average loss as your risk unit). Use it as your risk percentage in the position size calculator.

If the result is zero or negative, your numbers do not describe an edge, and Kelly says not to trade the strategy.

The formula

Kelly risk = p ÷ L − (1 − p) ÷ Wp is the win rate, W the average win and L the average loss, each in multiples of the amount risked. For W = L = 1 this is the familiar 2p − 1.
growth per trade = p × ln(1 + r × W) + (1 − p) × ln(1 − r × L)r is the risk per trade. The Kelly risk is the value of r that maximises this expression.

The Kelly formula for trading

The classic form is f* = p − q ÷ b, where p is the win rate, q = 1 − p, and b is the average win divided by the average loss. With a 55% win rate and winners 1.5 times the size of losers: 0.55 − 0.45 ÷ 1.5 = 25% of the account per trade. That figure assumes you know p and b exactly, which you never do.

Kelly with an uncertain edge: why traders use fractions

Your win rate is estimated from a limited sample, and Kelly is very sensitive to it. With 1.5R winners, the full Kelly size moves a lot across a plausible range of win rates:

Win rateFull KellyHalf Kelly
45%8.3%4.2%
50%16.7%8.3%
55%25.0%12.5%

Suppose you measure a 55% win rate and bet the full 25% Kelly size, but your true win rate is 45%. Your expected growth per trade at full Kelly is -1.49%, which is negative: the edge was overestimated, so a bet that looked optimal shrinks the account. At half Kelly the same mistake still grows (+0.39% per trade), and at quarter Kelly it grows slightly more (+0.48%). Fractional Kelly is insurance against estimation error.

Full, half and quarter Kelly vs a fixed 1% rule

Known edge: 55% win rate, 1.5R winners, 1R losers, 200 trades, 2,000 simulated accounts. Growth is the expected log growth per trade; drawdowns are medians and the 95th percentile of the worst peak-to-trough fall.

SizingRisk per tradeGrowth per tradeMedian max drawdown95th percentile drawdown
Full Kelly25.0%+4.57%92%99%
Half Kelly12.5%+3.44%65%84%
Quarter Kelly6.3%+2.02%39%56%
Fixed 1%1.0%+0.37%7%11%

Full Kelly maximises growth on paper but typically passes through a drawdown of about nine in ten of the account. Half Kelly gives up some growth for a smaller fall, and quarter Kelly smaller still. A fixed 1% grows slowly with a small drawdown. Most traders land between a quarter Kelly and a fixed small percentage, then size trades with the position size calculator and check survival with the risk of ruin calculator.

The Kelly formula calculator for betting: from odds to stake

A bet is the cleanest case of Kelly, because the payoff is fixed by the odds. At decimal odds d you win b = d − 1 for every 1 you stake, and lose the stake otherwise. The Kelly stake is the same expression as for trading with an average win of b and an average loss of 1:

stake = (b × p − q) ÷ bp is your estimate of the win probability, q = 1 − p and b = decimal odds − 1. It is the share of your bankroll to stake. A negative or zero result means the bet has no edge and the stake is zero.

A worked example, step by step

You estimate a 55% chance and the odds on offer are decimal 2.10.

  1. Net payoff. b = 2.10 − 1 = 1.10.
  2. What the odds imply. 1 ÷ 2.10 = 47.6%. Your estimate is higher, so you see an edge.
  3. Expected value per unit staked. 0.55 × 2.10 − 1 = +15.5%.
  4. Kelly stake. (1.10 × 0.55 − 0.45) ÷ 1.10 = 14.1% of the bankroll.
  5. Use a fraction. Half Kelly is 7.0% and quarter Kelly 3.5%. On a $1,000 bankroll that is $141, $70 and $35.

The same numbers are loaded in the calculator with the Bet: 55% at odds 2.10 preset. Add a bankroll to see the stake in money.

Converting odds to the input the calculator needs

The calculator takes decimal odds. Other formats convert like this: American positive odds +A give 1 + A ÷ 100; American negative odds −A give 1 + 100 ÷ A; fractional odds a/b give 1 + a ÷ b. The table shows common odds, the win probability they imply (the break-even point) and the Kelly stake if your own estimate were five points better.

OddsDecimalWhat the odds implyKelly if you are 5 points better (your probability)
decimal 1.501.50066.7%15.0% (at 72%)
decimal 2.00 (evens)2.00050.0%10.0% (at 55%)
American +1502.50040.0%8.3% (at 45%)
American −1101.90952.4%10.5% (at 57%)
fractional 5/23.50028.6%7.0% (at 34%)
American −2001.50066.7%15.0% (at 72%)

An edge buys a stake, and no edge buys none

Kelly is a function of the gap between your probability and the odds. At evens, a 50% estimate gives a stake of zero, and the stake grows quickly with the edge:

Your win probability at evens (decimal 2.00)Full Kelly stakeHalf Kelly
50%0.0%0.0%
52%4.0%2.0%
55%10.0%5.0%
60%20.0%10.0%
65%30.0%15.0%

That sensitivity is the weak point. The input that matters most, your own win probability, is the one you know least precisely, which is why the calculator is only as good as your estimate.

Kelly for trading versus betting

  • The payoff. A bet has one fixed payoff and a known stake. A trade has a distribution of outcomes, a stop that may not fill where you set it, gaps and costs. Reduce a strategy to an average win, an average loss and a win rate, and treat the result as an approximation.
  • The probability. In betting you assign a probability and compare it with the odds. In trading you estimate it from a trade history, which changes as the market does.
  • What the stake means. In betting the Kelly fraction is the money you stake. In trading it is the share of the account you lose if the stop is hit, so you turn it into a position size with the position size calculator.
  • Independence. Kelly assumes each bet is independent and sized one at a time. Several correlated bets or positions open at once carry more risk than the sum of their separate Kelly stakes, so size down.

Why full Kelly is aggressive

Kelly maximises the average logarithm of wealth, which is the long-run growth rate if your inputs are exactly right. It says nothing about the ride. Take the case above at evens with a true 55% win probability, so full Kelly is 10% of the bankroll, simulated over 200 bets with 4,000 runs:

StakeShare of bankrollGrowth per bet (average log)Median worst drawdownChance of falling below half within 200 bets
Full Kelly10.0%+0.50%70%38%
Half Kelly5.0%+0.38%42%8%
Quarter Kelly2.5%+0.22%23%0%
Twice Kelly20.0%-0.01%96%79%

Full Kelly has the highest growth, and in this simulation about 38% of the runs fall below half the bankroll at some point. Half Kelly keeps 75% of the growth, and the chance of halving the bankroll falls to 8%. Quarter Kelly is calmer still (0.3%). Staking twice Kelly has no positive growth at all, and about four in five runs touch a 50% loss.

When your estimate is wrong

Suppose you believe the win probability is 55% at evens, so you stake the full 10%, but the true probability is 50% and the bet has no edge. The growth per bet is then -0.50%, negative, and it stays negative at half Kelly (-0.13%) and quarter Kelly (-0.031%), only much smaller. A fraction of Kelly turns an estimation error into a slow loss instead of a fast one. Betting and trading both carry a risk of loss, and Kelly cannot create an edge that is not there.

Link to risk of ruin and drawdown

The table above is a drawdown question: how deep does the account fall on the way. The risk of ruin calculator answers the same question for your own numbers, and the drawdown recovery calculator shows why a 50% fall needs a 100% gain to recover. The guides what is risk of ruin and how much to risk per trade cover the reasoning in plain language. If betting stops being fun, help is available from the gambling support services in your country. This page is for education, not advice.

Worked examples

Even odds with a small edge

You win 55% of trades with winners and losers the same size. Full Kelly is 2 × 0.55 − 1 = 10% per trade. Half Kelly is 5% and quarter Kelly is 2.5%. Simulating 200 trades shows the typical worst drawdown:

70% drawdownFull Kelly (10%)
42% drawdownHalf Kelly (5%)
23% drawdownQuarter Kelly (2.5%)

Low win rate, big winners

You win 40% of trades at 3R. Kelly risk is 0.4 − 0.6 ÷ 3 = 20%, a number far higher than most traders could stomach. That is the main reason to use a fraction of Kelly.

20% per tradeFull Kelly
10%Half Kelly
5%Quarter Kelly

When Kelly says no

At a 40% win rate with winners and losers the same size, Kelly is 2 × 0.4 − 1 = −20%. A negative result means the strategy has negative expectancy, and no position size will fix it.

−20%Kelly
Do not trade thisVerdict

A bet at decimal odds 2.10 with a 55% estimate

The odds imply 47.6%, you estimate 55%. Full Kelly is 14.1% of the bankroll, half Kelly 7.0%. The expected value is +15.5% per unit staked.

14.1%Full Kelly
7.0%Half Kelly
3.5%Quarter Kelly

No edge at evens

At decimal 2.00 the odds imply 50%. A 50% estimate gives a Kelly stake of zero, and a 52% estimate only 4%. The stake is the edge, not the odds.

0%At 50%
4%At 52%
20%At 60%

Frequently asked questions

Should I bet full Kelly?

Rarely. Full Kelly maximises growth only if your inputs are exact, and it produces large drawdowns. Half or quarter Kelly keeps most of the growth with far less risk.

What is half Kelly?

Half Kelly means risking half the amount the Kelly formula suggests. It gives about 75% of the growth with a much smaller drawdown, which is why many professionals prefer it.

How do I use the Kelly number as a trader?

Treat it as an upper limit for risk per trade, then choose a fraction of it. Enter the result as your risk percentage in the position size calculator.

What happens if I bet more than Kelly?

Growth falls. At about twice Kelly it reaches zero, and beyond that the account shrinks over time even with a profitable strategy.

Is the Kelly criterion the same for stocks and sports betting?

The formula for a binary outcome is the same. Trading is harder because win rate and payoff are uncertain and change. That is another reason to use a fraction.

Where do my win rate and payoff come from?

From your own trade history. You need enough trades for the numbers to be reliable, ideally a few hundred. With less data, use a lower win rate to be safe.

Is Kelly safe for trading?

Full Kelly is not safe for most traders. It maximises long-run growth only if your win rate and payoff are known exactly, and it produces very large drawdowns even then. Overestimating your edge can turn the bet negative. Most traders use half or a quarter of Kelly, or a small fixed percentage.

Why half Kelly?

Half Kelly keeps most of the growth while cutting the drawdown sharply, and it tolerates errors in your estimated edge. The tables on this page show the trade-off.

Does the Kelly criterion work for crypto, forex or stocks?

The formula works in any market given a reliable win rate and payoff. The practical problem is estimating them: markets change, and short samples are noisy. That is the main reason to use a fraction of Kelly.

What if my win rate changes?

Then the optimal size changes with it. Recalculate when your sample grows, use a conservative estimate (for example the lower end of your measured range), and keep the fraction small enough that a worse true win rate does not make the bet negative.

How do I calculate the Kelly stake from decimal odds?

Subtract 1 from the decimal odds to get b, then use (b × p − q) ÷ b, where p is your win probability and q = 1 − p. At decimal odds 2.10 and a 55% win probability the full Kelly stake is 14.1% of the bankroll.

What is the Kelly formula for betting?

The stake as a share of the bankroll is (b × p − q) ÷ b, with b the net odds (decimal odds − 1), p the win probability and q = 1 − p. It is the same formula as for trading, with an average win of b and an average loss of 1. A negative result means no bet.

How do I use the Kelly formula with American odds?

Convert to decimal first: +150 becomes 2.50 (1 + 150 ÷ 100) and −110 becomes 1.909 (1 + 100 ÷ 110). Then enter the decimal odds and your win probability. The conversion table on this page lists the usual ones.

Is Kelly betting profitable?

Kelly maximises long-run growth only if you have a real edge and your probability is accurate. It does not create an edge. If your estimate is too high, even a fraction of Kelly loses slowly. Treat the result as a ceiling and use half or quarter Kelly.

What is the difference between Kelly and flat staking?

Flat staking risks the same amount every time, so it ignores the size of the edge. Kelly stakes more when the edge is larger and less when it is smaller, and stakes nothing with no edge. The cost is much larger swings: full Kelly in the example above has a 38% chance of halving the bankroll within 200 bets.

Why does the Kelly criterion use the logarithm of wealth?

Wealth compounds, so a 50% loss followed by a 50% gain leaves you behind. Maximising the average logarithm maximises the compounded growth rate, and it automatically penalises large losses. That is why growth falls past the Kelly point and why twice Kelly has no growth at all.

Kelly sizing depends on a reliable edge, so read the Kelly criterion for trading for the formula and its limits, and what expectancy is for how to measure that edge.

Guides that explain it

a.b

Built by a.b. I joined TradingView in December 2015 and publish indicators there. These tools use the same standards: clear logic, no hidden assumptions. Message me on TradingView chat with feedback or a tool you would like to see.

Last updated November 2026. For education only. Not financial advice. Results depend on your inputs and cannot predict future performance.