What Is the Kelly Criterion and Should Traders Use It?
Direct answer
Kelly criterion in practice
The Kelly criterion is a formula for the fraction of capital to risk on a bet with a known edge: f = W − (1 − W) ÷ R, where W is the win probability and R the win-to-loss ratio. With a 55% win rate and 1:1 payoffs it says risk 10% per bet; with 40% and 2:1 payoffs, also 10%. In real trading the inputs are uncertain and overestimating them leads to huge drawdowns, so most practitioners use half Kelly or much less.
- Kelly fraction f = W − (1 − W) ÷ R
- Example: W = 55%, R = 1 gives f = 10% of capital per bet
- Full Kelly maximises long-run growth but brings deep drawdowns
- Half Kelly keeps about three quarters of the growth with much less volatility (for a known edge)
The Kelly criterion has a cult following among quantitative traders and prediction-market bettors. It answers a precise question, but only if you can answer two other questions precisely first.
What people on X are asking
In October 2025 a prediction-market user asked what the Kelly strategy is and how it can help you trade smarter, noting that pro traders mention it often. In March 2026 a reply asked an investor whether they allocate full Kelly or half Kelly, and in September 2025 someone asked another trader if he knew about the half Kelly criterion. Here is the formula, an honest look at its inputs, and why "half" is so common.
The fact-checked answer
Origin. John L. Kelly Jr., a Bell Labs scientist, published the idea in 1956 in the Bell System Technical Journal, framing it in terms of information theory. It finds the bet size that maximises the expected logarithm of wealth, which is equivalent to the fastest long-run compound growth.
The trading form. For a bet that wins R times the stake with probability W and loses the stake with probability 1 − W, the optimal fraction is f = W − (1 − W) ÷ R. If f is zero or negative, there is no edge and the formula says bet nothing.
Kelly fraction for different edges
Computed by UXS from f = W − (1 − W) ÷ R. In trading, "fraction" means the share of capital lost if the stop is hit, not the position size. Tap a column header to sort.
| 50% | 1.0 | 0% | 0% | 0% |
| 55% | 1.0 | 10.0% | 5.0% | 2.5% |
| 60% | 1.0 | 20.0% | 10.0% | 5.0% |
| 40% | 2.0 | 10.0% | 5.0% | 2.5% |
| 35% | 2.0 | 2.5% | 1.25% | 0.6% |
| 30% | 3.0 | 6.7% | 3.3% | 1.7% |
| 45% | 1.5 | 8.3% | 4.2% | 2.1% |
Why traders scale it down. Full Kelly is aggressive even when the edge is known exactly: it accepts frequent drawdowns of 50% or more on the way to maximum growth. Standard results in the literature on fractional Kelly show that betting half the Kelly amount gives about 75% of the long-run growth rate with roughly half the volatility. Halving the bet costs a quarter of the growth: a good trade.
The real problem: you do not know W and R. Your win rate from 50 trades might be 55%, but the true long-run figure could easily be 48%. Plug 55% into Kelly when the truth is 50% and you are betting 10% per trade on a strategy with no edge, a fast way to lose most of the account. Because overbetting hurts much more than underbetting, uncertainty pushes the sensible size well below the formula's answer.
Compare with fixed-fraction risk. Many traders risk 0.5–2% per trade. In Kelly terms that is a small fraction of the full Kelly figure for typical retail edges, which is a reasonable response to uncertainty. Our page on how much to risk per trade covers that approach, and win rate vs expectancy shows how to estimate your inputs.
To compare strategies on risk-adjusted terms, see what a good Sharpe ratio is.
Decision helper
| Situation | Kelly fits? | Better approach |
|---|---|---|
| Known odds (casino-style, simple bets) | Yes, the original use case | Fractional Kelly still safer |
| Trading strategy with 30 recorded trades | Poorly: inputs too uncertain | Fixed small risk % |
| Long track record, stable edge | Partly | Quarter to half Kelly as a ceiling |
| Several correlated positions | No (formula assumes one bet at a time) | Treat as one combined bet |
| Prop-firm challenge with a drawdown limit | No | Size from the drawdown limit instead |
Checklist before using Kelly
- Estimate W and R from a large sample of your own trades.
- Shrink the estimates towards breakeven to allow for luck.
- Compute full Kelly, then use a quarter to a half as a maximum.
- Never exceed a hard cap such as 2% risk per trade.
- Recalculate as more data come in.
- Combine correlated trades before applying the formula.
Quick poll
How do you size your trades?
What each answer means
- Full Kelly: Mathematically fastest growth only if your inputs are exactly right; in practice drawdowns can be severe.
- Half or quarter Kelly: A common compromise: most of the growth with far smaller swings, still sensitive to input errors.
- Fixed % of the account per trade: Simple and robust; it ignores differences in edge between setups.
- Fixed position size: Easy, but risk changes with stop distance and account size.
- No method yet: Start with a small fixed % and record trades so you can measure your edge.
Your answer is stored only in this browser (localStorage). UXS does not collect, count or publish votes, so no totals are shown.
Common mistakes
- Using a win rate from a short, lucky streak.
- Applying Kelly to position size instead of the amount at risk.
- Ignoring correlations across open trades.
- Forgetting that Kelly maximises growth, not comfort or survival under a drawdown limit.
FAQ
What is the Kelly criterion formula?
f = W − (1 − W) ÷ R, where W is the probability of winning and R is the average win divided by the average loss.
What is half Kelly?
Betting half the fraction the formula suggests. For a known edge it keeps about three quarters of the growth with much lower volatility.
Is the Kelly criterion good for trading?
As a ceiling and a way of thinking, yes. As an exact sizing rule, it is dangerous because real win rates and payoffs are uncertain.
What if the Kelly result is negative?
The strategy has no edge on those inputs, so the formula says not to bet.
Who invented the Kelly criterion?
John L. Kelly Jr. at Bell Labs, in a 1956 paper.
Does Kelly work for prediction markets?
It is often used there because each bet has a stated probability, but your estimate of the true probability still has to be better than the market's.
Sources
- J. L. Kelly Jr., A New Interpretation of Information Rate (Bell System Technical Journal, 1956)
- SEC Investor.gov: Risk tolerance
- FINRA: Day trading
- CFTC: Futures market basics
Facts checked November 10, 2026. X posts are cited as evidence of the question, not of the answer.