Win rate (%)
Risk:Reward ratio
Account balance ($, optional)
Result (click to copy)
Choosing Your Kelly Fraction
Half-Kelly is the standard professional default
It captures roughly 75% of Full Kelly's growth rate while cutting variance substantially — this is what most professional traders and bettors actually use, not Full Kelly.
Go smaller than Half-Kelly if your edge estimate is uncertain
Kelly assumes your win rate and R:R are exactly correct. If they're based on a small sample or a hunch rather than a solid trading history, Quarter-Kelly or less is more appropriate.
A negative Kelly means stop, not size down
If Full Kelly is negative, no position size — however small — makes the strategy profitable on average. Fix the edge first.
How to Use the Kelly Criterion Calculator
- Enter your win rate — use historical data from at least 50 trades. Kelly is sensitive to win rate accuracy — overestimating by 5% can significantly change the recommended size.
- Enter your R:R ratio — the average reward relative to risk across your winning trades. If you target 15 pips with a 10 pip stop, enter 1.5.
- Enter account balance — optional but enables the calculator to show exact dollar amounts alongside percentages.
- Read the Kelly outputs — the calculator shows Full Kelly, Half-Kelly, and Quarter-Kelly percentages with explanations of when to use each.
- Check the sensitivity table — see how Kelly changes at different win rates to understand how sensitive your optimal size is to win rate accuracy.
- Apply to position sizing — use the Half-Kelly percentage as your risk per trade input in the Position Scaling Calculator for complete trade sizing.
⚠️ Important: Never use Full Kelly in live trading. It creates extreme variance — drawdowns of 30–50% are mathematically expected even with a genuine edge. Half-Kelly or Quarter-Kelly is the professional standard.
Understanding the Kelly Criterion
The Kelly Criterion was developed by John L. Kelly Jr. at Bell Labs in 1956 and later adopted by gamblers and investors to calculate the optimal fraction of capital to risk on each bet. Ed Thorp — the mathematician who beat blackjack and later the stock market — popularised its use in financial markets.
📐 The Formula
Kelly % = Win Rate − (Loss Rate ÷ Reward Ratio). Example: 55% WR, 1:1.5 RR → Kelly = 0.55 − (0.45÷1.5) = 0.55 − 0.30 = 0.25 = 25% per trade.
✅ Positive Kelly
Any positive Kelly value means you have a mathematical edge. The higher the Kelly %, the stronger your edge. Negative Kelly means you have no edge — do not trade this strategy with real money.
⚖️ Half-Kelly
Half the Full Kelly amount. Gives approximately 75% of the growth rate with dramatically lower variance. Preferred by professional traders and quantitative funds.
🛡️ Quarter-Kelly
25% of Full Kelly. Very conservative — similar growth to Half-Kelly but with minimal drawdown. Suitable for new strategies where win rate confidence is low.
📊 Growth Maximisation
Full Kelly mathematically maximises the long-run growth rate of capital. However, it also maximises short-term volatility, making it psychologically and practically challenging to stick to during losing streaks.
🔄 Kelly Rebalancing
After each trade, the Kelly-optimal size changes as your balance changes. Fixed fractional risk (which is effectively rebalanced Kelly) achieves this automatically.
Kelly Criterion in Practice
Why traders underperform Kelly
Most traders who attempt to use Full Kelly abandon it after the first significant losing streak. A 25% Kelly with a string of 6 losses can reduce a $10,000 account to $5,000 — mathematically expected but emotionally devastating. Half-Kelly produces slower growth but the drawdowns are survivable, allowing traders to maintain discipline through variance.
Kelly and prop firms
Kelly Criterion is most useful for sizing trades in personal accounts without strict drawdown limits. For prop firm challenges, Kelly should be used as a reference point — then reduced significantly. A Kelly of 8% should translate to 0.5–1% actual risk in a prop firm context where a 10-consecutive-loss streak would breach the account limit.
Kelly with correlated positions
Kelly assumes each bet is independent. In trading, many positions are correlated (especially in forex where USD movement affects multiple pairs simultaneously). When running multiple correlated positions, divide the Kelly size by the number of correlated positions to maintain the intended total exposure.
🎯 Practical application: Use Kelly to confirm you have a positive edge (any positive Kelly%) and to establish a maximum size ceiling. Then use Fixed Fractional at 0.5–1% for actual prop firm trading. Kelly tells you what's mathematically optimal — risk management tells you what's practically sustainable.
Risk of Ruin & Sample Size
💀 What "risk of ruin" means
Risk of ruin is the probability your account falls so low you effectively can't continue — commonly defined as dropping below about 10% of your starting balance, since recovering from there requires a 10x gain just to get back to even. Even a genuinely positive edge doesn't reduce this to zero; it only makes it low.
🎲 Full Kelly carries real ruin risk even with an edge
At Full Kelly, a 50% drawdown from peak happens roughly once per year of normal trading, even when your edge estimate is exactly correct. Half-Kelly cuts that to roughly 25%, Quarter-Kelly to roughly 12% — this is the real, quantified reason professionals size down from Full Kelly, not just caution for its own sake.
📉 Small samples lie
The math behind Kelly is exact, but your win rate and R:R inputs come from your trading history — and 20-30 trades is not enough to know your true win rate with any confidence. If your Kelly number is built on a small sample, treat it as a hypothesis to keep testing, not a precise number to size by.
Avoid These Mistakes
Sizing at Full Kelly in real trading
Mathematically optimal for long-run growth, but the drawdowns are real and frequent — almost no trader can psychologically tolerate them, leading to abandoning the system at exactly the wrong time.
Trusting a Kelly number from too few trades
Kelly is only as good as your win rate and R:R inputs. A win rate from 20-30 trades can easily be off by 10+ percentage points from your true long-run rate — use a conservative fraction until your sample is larger.
Applying Kelly to correlated positions independently
Kelly assumes each bet is independent. Running several correlated positions (e.g. multiple USD pairs) at full Kelly size each effectively multiplies your real risk beyond what any single Kelly calculation accounts for.
Frequently Asked Questions
What is the Kelly Criterion?
The Kelly Criterion is a mathematical formula that calculates the optimal fraction of your capital to risk on each trade to maximise long-term growth: Kelly% = Win Rate − (Loss Rate ÷ Risk:Reward). A positive Kelly% means you have a positive edge; negative means you don't.
Why use Half-Kelly instead of Full Kelly?
Full Kelly maximises long-term growth mathematically, but creates massive drawdowns in practice — sometimes 50%+ swings. Half-Kelly gives approximately 75% of the growth with dramatically lower variance. Most professional traders and quant funds use fractional Kelly (25-50%) for this reason.
What does a negative Kelly mean?
A negative Kelly means your strategy has no mathematical edge — you're expected to lose money over time. This happens when your win rate is too low for your R:R ratio (or vice versa). Fix: increase your win rate, increase your R:R, or both.
Is Kelly Criterion suitable for prop firm trading?
Yes, but use conservative fractions (Quarter-Kelly or less). Prop firms have strict drawdown limits that punish large position sizes during losing streaks. The Kelly Criterion tells you your mathematical edge — combine it with your firm's daily drawdown limit to set your actual risk per trade.
References
📖 J. L. Kelly Jr., "A New Interpretation of Information Rate" (1956)
The original Bell Labs paper the Kelly Criterion is named after — foundational reading for understanding where the formula comes from.
📊 Your own trading journal
Kelly is only as accurate as the win rate and R:R you feed it — a real, logged trading history of 50+ trades gives a far more trustworthy input than an estimate or a short recent streak.