Kelly Criterion: Optimal Position Sizing Formula
⚡ Read this before you open your next trade
The Kelly Criterion is a mathematical formula for determining optimal position size to maximize long-term geometric growth rate given a known edge. Developed by Bell Labs scientist John Kelly in 1956 and popularized by gambling mathematician Edward Thorp, Kelly has become the gold standard for position sizing in systems with measurable edge. The formula: f* = (bp - q) / b where b = odds (reward-to-risk ratio), p = win probability, q = loss probability (1-p), f* = optimal fraction of capital to risk. Example: 60% win rate, 2:1 reward-to-risk → f* = (2 × 0.6 - 0.4) / 2 = 0.4 = 40% of account per trade. This sounds enormous, which is why most traders use fractional Kelly (25-50% of the calculated value) to reduce volatility.
Why Kelly Maximizes Geometric Growth
Consider a 60% win rate with 2:1 payoff. Common intuition says: "bet big because I win more than I lose." But position size has a mathematical ceiling. Too small and you don't capitalize on your edge. Too large and a losing streak crushes you (going to 50% drawdown requires 100% gain to recover, 75% drawdown requires 400% gain). Kelly finds the exact size that maximizes expected LOG of return — the geometric mean of per-trade returns. Why log? Because wealth compounds multiplicatively. A sequence of +10% then -10% leaves you at 99% (not 100%). Kelly accounts for this "volatility tax." At f*, geometric growth per trade peaks; betting more or less than f* reduces long-term compound growth. Full Kelly is the provably optimal strategy for known-edge bets assuming infinite betting opportunities and continuous rebetting.
The Kelly Formula in Practice
Applied calculation for trading. Suppose your strategy shows: 55% win rate, average winner 80 pips, average loser 60 pips over a 200-trade sample. Kelly inputs: p = 0.55, q = 0.45, b = 80/60 = 1.33. f* = (1.33 × 0.55 - 0.45) / 1.33 = (0.7315 - 0.45) / 1.33 = 0.2816 / 1.33 = 21.2%. Full Kelly says risk 21% of account per trade. Half Kelly (recommended for most): 10.6%. Quarter Kelly: 5.3%. Compare to 1% fixed fractional — Kelly suggests 5-20x larger position sizes for this edge. The difference in compound growth over years is enormous: a $10k account with 1% risk per trade might grow to $40k over 3 years; the same edge with half-Kelly sizing grows to $200k. But the path volatility is much higher — Kelly can produce 40-50% drawdowns that 1% sizing never approaches.
Why Full Kelly Is Too Aggressive
Full Kelly has three real-world problems that make fractional Kelly preferable. (1) Edge uncertainty — Kelly assumes you KNOW win rate and payoff exactly. In trading, you estimate from samples. If your "55% win rate" was based on 200 trades, the 95% confidence interval is ±7% — actual edge might be 48-62%. Using full Kelly with overestimated edge causes over-leveraging and potential blow-up. (2) Psychological tolerance — full Kelly produces 40-50% drawdowns as a normal feature. Most humans cannot psychologically handle seeing their account cut in half even when mathematically it's optimal. Emotional decisions during drawdowns destroy systematic advantages. (3) Edge decay — markets change; your edge may decline or disappear. If you've been sized at full Kelly for a now-vanished edge, you're over-leveraged for the new (non-edge) reality. Fractional Kelly provides safety margin for these realities.
⚠️ Mistake most traders make
Reading about trading is not enough. Traders who practice in real time — tracking signals, analyzing their trades, and learning from results — improve 3x faster. In the Take Profit app, you can do this right away.
Half-Kelly and Quarter-Kelly
Practical recommendations: Half-Kelly (50% of full Kelly) preserves about 75% of the theoretical geometric growth rate while reducing expected maximum drawdown from ~40% to ~25%. The sharp improvement in drawdown for small growth sacrifice makes half-Kelly popular among professional money managers. Quarter-Kelly preserves ~50% of theoretical growth with ~12% max drawdown — suitable for traders with lower risk tolerance or unstable edges. Fractional Kelly also provides some protection against edge estimation error — if true edge is 20% lower than estimated, full Kelly is over-sized but half-Kelly is close to full Kelly for the true edge. Rule of thumb for most retail traders: start with quarter-Kelly until you've proven the edge over 500+ trades across multiple market regimes. Advance to half-Kelly only after robust out-of-sample validation.
Kelly Implementation Challenges
Several practical problems complicate Kelly implementation. (1) Variable payoffs — Kelly formula assumes fixed reward:risk per trade, but trading has variable outcomes (stops might be at different distances, targets vary). Solutions: use average win/average loss over recent sample, or use "generalized Kelly" for continuous distributions (more complex math). (2) Correlated positions — Kelly assumes independent bets. If you have multiple simultaneous positions with correlated outcomes (e.g., long EUR/USD and short USD/JPY are correlated), aggregate risk is higher than individual Kelly calculations suggest. Scale down accordingly. (3) Estimation error compounds — small errors in win rate estimation translate to large position sizing errors at full Kelly. A 5 percentage point overestimate of win rate approximately doubles recommended position size. (4) Regime shifts — edges that worked in 2020-2022 may not work in 2026; Kelly computed on old data is wrong for current reality. Continuous monitoring and regular re-estimation (e.g., rolling 200-trade window) helps but doesn't eliminate this problem.
💡 Most traders read this and... do nothing
Want to see this on a live market?
Reading is 10% of learning. The other 90% is watching a real market. In the Take Profit app, you see how theory works in practice — every day.
- Signals with entry, SL, TP — and the result (73% win rate)
- Trading journal — log every trade and learn from mistakes
- Macro calendar — know when NOT to trade
- AI analysis — understand what the market says today
Related Guides
Position Sizing Explained
Master position sizing in trading. Learn how to calculate the right lot size for every trade based on your account balance, risk tolerance, and stop loss distance.
The Risk Per Trade Rule
Understand the 1% and 2% risk per trade rules used by professional traders. Learn how to set your risk limit and why it protects your trading capital.
Martingale System: Why Doubling Down Destroys Accounts
Understand the martingale betting system — mathematical inevitability of ruin, why it feels successful in short term, and safer alternatives for traders.
Maximum Drawdown in Trading
Learn what maximum drawdown is and why it matters. Understand how to measure, manage, and reduce drawdowns to protect your trading account long-term.
Risk of Ruin: Probability of Going Broke
Calculate your risk of ruin — formula explanation, inputs (win rate, payoff, risk per trade), interpretation, and how to reduce blow-up probability to effectively zero.
→Sound familiar?
•"You enter a trade and instantly regret it"
•"You don't know why the market moved — again"
•"You copy signals but don't understand the reasoning"
•"Trading feels like guessing"
It's not about intelligence — it's about tools. See what trading with structure looks like.
Frequently Asked Questions
What win rate do I need for Kelly to be positive?
Depends on payoff ratio. Break-even Kelly at p×b = q, i.e., p = 1/(1+b). With 2:1 payoff, you need p > 33% to have positive Kelly (positive edge). With 1:1 payoff, you need p > 50%. With 3:1 payoff, p > 25%. If Kelly calculates negative, don't trade that setup — you have no edge. Kelly automatically filters out losing strategies by returning zero or negative values.
Can I use Kelly on a small account?
Mathematically yes, but practically challenging. Kelly's effectiveness depends on compound growth, which requires thousands of trades to manifest fully. Small accounts also face minimum lot sizes that make fractional sizing difficult. Additionally, psychological impact of Kelly-sized drawdowns on small accounts can be severe — losing $5k on a $10k account is emotionally different from losing $500k on a $1M account. For small accounts, fixed fractional (1% risk) is often more practical than Kelly even if theoretically suboptimal.
How does Kelly handle multiple concurrent positions?
Standard Kelly assumes single bet at a time. For portfolio Kelly with multiple positions, you need covariance-adjusted calculation. Simplified approach: if positions are highly correlated (correlation > 0.7), treat them as one combined position and compute Kelly on aggregate. If positions are uncorrelated (correlation < 0.3), sum Kelly sizes up to total portfolio heat limit (typically 10-20% of full Kelly per position, 40-60% total). Never run full Kelly on each of many correlated positions — aggregate risk exceeds Kelly-optimal levels quickly.
What's the difference between Kelly and Optimal f?
Kelly formula is for binary outcomes (win or lose with fixed amounts). Ralph Vince's "Optimal f" generalizes Kelly to continuous distributions — it calculates optimal fractional exposure given the full distribution of trade outcomes, not just win/loss. Optimal f is theoretically more accurate for real trading where win/loss amounts vary, but it requires more data and more computation. In practice, half-Kelly on average win/loss gives results close to Optimal f while being simpler to calculate.
Do hedge funds use Kelly?
Systematic hedge funds often use Kelly-based sizing with significant fractional adjustments — typically quarter-Kelly or even eighth-Kelly given their long time horizons and low drawdown tolerance (investors redeem during drawdowns, which destroys compound growth). Famous Kelly proponents include Jim Simons' Renaissance Technologies, Ed Thorp's Princeton Newport Partners, and many CTAs. Most discretionary hedge funds don't use explicit Kelly formulas — they use rule-of-thumb position sizing (1-3% per trade) that conservatively approximates fractional Kelly.
Why trust us
Active trader since 2020
Actively trading financial markets since 2020.
Thousands of users
A trusted community of traders using our analysis daily.
Real market analysis
Daily analysis based on data, not guesswork.
Education, not advice
Transparent educational content — you make the decisions.
About the author
Kacper MrukXAUUSD & ETHUSD Trader | Macro + options data | Think, don't follow
Creator of Take Profit Trader's App. Specializes in XAUUSD and ETHUSD, combining macro analysis with options data. He teaches not how to trade, but how to think in the market. Actively trading since 2020.
Related Topics
Before you download — check yourself:
Start free