This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

This is one of those topics where the conventional wisdom doesn’t quite hold up.

The Kelly Criterion: The Mathematical Formula for Optimal Position Sizing

Understanding how to size your investment positions using probability and expected value can transform your long-term returns and reduce catastrophic losses.

Here’s the thing most people miss about this topic.

Introduction: Why Position Sizing Matters

Most investors focus extensively on where to invest their money—which stocks to buy, which sectors to favor, which market conditions signal opportunity. Yet one of the most critical decisions receives far less attention: how much to invest in each position. Position sizing determines not just your returns, but your survival as an investor.

Related: index fund investing guide

Legendary investors have long understood this principle. Warren Buffett allocates concentrated positions to his highest-conviction ideas. George Soros built his wealth partly through dynamic position sizing based on probability estimates. But what if there were a mathematical formula to guide these decisions? What if you could quantify the optimal amount to risk on any given trade?

This is where the Kelly Criterion comes in. Developed by John L. Kelly Jr. in 1956, this elegant mathematical formula calculates the exact percentage of your bankroll to wager on any bet with known probability and payoff characteristics.¹ The formula has applications far beyond gambling—it’s become increasingly important for serious investors seeking to maximize long-term wealth accumulation while managing downside risk.

The Kelly Criterion Formula Explained

The Kelly Criterion formula, in its most common form for investment applications, is expressed as:

Where:

• f* = The fraction (or percentage) of your bankroll to allocate to the position
• b = The odds received on the bet (the ratio of profit to stake)
• p = The probability of winning (as a decimal)
• q = The probability of losing (equals 1 – p)

For a simpler binary outcome scenario (where you either win or lose a fixed amount), the formula becomes:

Or even more intuitively:

The beauty of this formula lies in its mathematical elegance. It’s not arbitrary or based on gut feeling. Rather, it emerges from the mathematics of logarithmic utility—specifically, the formula maximizes the expected value of the logarithm of your wealth over time, which directly translates to the highest compound annual growth rate (CAGR) for your portfolio.²

The Mathematics Behind Kelly

Understanding why the Kelly Criterion works requires a brief excursion into probability and compounding. Consider a series of bets where you repeatedly risk capital and receive returns based on successful outcomes. Your cumulative wealth over n bets grows as:

Wealth = Initial Capital × (1 + f × return₁) × (1 + f × return₂) × … × (1 + f* × returnₙ)

The geometric mean of these returns—not the arithmetic mean—determines your long-term wealth. The Kelly Criterion maximizes this geometric mean growth rate. In practical terms, this means it provides the highest probability of having more wealth in the future compared to any other position-sizing strategy.

The derivation involves maximizing the expected logarithmic utility function. When you take the logarithm of your wealth and maximize its expected value across many bets, the solution—solved through calculus—yields precisely the Kelly formula. This mathematical foundation gives the criterion its power: it’s not a heuristic; it’s a proven optimal solution to a defined mathematical problem.³

Practical Example: Applying Kelly to a Stock Investment

Let’s work through a concrete example to see how Kelly sizing works in practice.

The Kelly Criterion in Portfolio Management

Professional investors rarely use the Kelly Criterion at full strength. Instead, many employ a “fractional Kelly” approach, using 25-50% of the recommended Kelly fraction. Why?

Full Kelly optimization assumes perfect knowledge of probabilities and payoffs. In reality, estimate errors are inevitable. If you overestimate your winning probability by even a few percentage points, full Kelly sizing can lead to catastrophic losses. The formula is mathematically elegant but practically fragile when inputs are uncertain.

Fractional Kelly offers insurance: Using half-Kelly (betting 50% of the calculated fraction) dramatically reduces the impact of estimation errors while sacrificing only modest growth rates. Using quarter-Kelly provides even more safety. Research shows that experienced investors often find optimal risk-adjusted returns using 25-50% Kelly sizing rather than full Kelly.

Consider that Ed Thorp, the mathematician who brought Kelly’s work to financial markets, used Kelly-based sizing but typically applied only a fraction of the theoretical maximum. His hedge fund generated 20% annual returns with relatively low volatility for decades—suggesting that fractional Kelly applied with precision beats full Kelly with estimation error.²

Calculating Win Probability: The Critical Assumption

The Kelly formula’s greatest strength is also its greatest weakness: it requires you to estimate probabilities. How do you know your 60% win probability is actually 60% and not 50% or 70%?

This question has no perfect answer. However, serious investors use several approaches:

Historical Win Rate Analysis

If your investment approach has a documented track record, calculate your actual historical win rate. If you’ve executed similar trades 50 times and won 65% of them, 65% is your empirical p value. This works best for systematic, repeatable strategies where historical conditions resemble future conditions.

Scenario Analysis and Statistical Modeling

For individual positions, develop a range of scenarios (bull case, base case, bear case) with associated probabilities, then back into a weighted probability of success. Use historical volatility and correlations as inputs to statistical models of expected returns.

Bayesian Updating

Start with a prior probability based on historical data, then update your estimate as new information arrives. This approach combines quantitative analysis with real-time market information.

Conservative Estimation

Deliberately estimate probabilities below what your analysis suggests. If modeling indicates 65% win probability, use 60%. This built-in margin of safety protects against overconfidence bias.

When Kelly Breaks Down: Important Limitations

The Kelly Criterion is powerful but not universal. Several conditions limit its application:

Non-Binary Outcomes

Kelly assumes binary outcomes (you either win or lose). Real markets produce a spectrum of returns. Extending Kelly to continuous return distributions requires more sophisticated mathematics, though modified versions exist for normal distributions.

Reinvestment Assumptions

Kelly assumes you reinvest gains and losses back into the next wager. If you need to withdraw capital for living expenses or other purposes, Kelly sizing may be suboptimal.

Correlation Effects

Kelly works for independent bets. In a portfolio of correlated positions, the mathematics becomes more complex. Using Kelly independently for each position ignores how positions interact—potentially leading to excessive overall portfolio risk.

Unknown or Changing Probabilities

As mentioned, Kelly requires probability estimates. For truly novel situations or black-swan events, probability estimation becomes nearly impossible. In these cases, prudent investors revert to much smaller position sizes.

Liquidity Constraints

Kelly might recommend a 40% position in an illiquid micro-cap stock. But if the stock becomes hard to exit, you can’t reduce your position quickly when conditions change. Position size must also respect market liquidity and your ability to trade.

Kelly Criterion vs. Other Position Sizing Methods

How does Kelly compare to alternative approaches?

Fixed Dollar Amount

Many traders risk a fixed dollar amount per trade regardless of probability or payoff characteristics. This ignores information about edge and is mathematically suboptimal. However, it’s simple and prevents massive position blowouts.

Percentage Risk (1-2% Rule)

Risking a fixed percentage of your portfolio on each trade (commonly 1-2%) is conservative but ignores the quality of each opportunity. A high-probability, favorable-payoff trade gets the same allocation as a lower-conviction idea.

Volatility-Based Sizing

Position size inversely to expected volatility, ensuring that whether you invest in low-volatility bonds or high-volatility stocks, you maintain consistent portfolio volatility. This is practical for diversified portfolios but ignores differences in expected return.

Kelly Criterion

Allocates proportionally to edge (expected return) relative to risk. Mathematically optimized for long-term wealth maximization given accurate probability estimates. Requires careful probability estimation and fractional application in practice.

Verdict: For systematic traders with reliable historical data on their edge, Kelly (especially fractional Kelly) typically outperforms simpler methods. For casual investors with sparse historical data, simpler percentage-risk methods may be safer due to lower estimation error risk.

Implementing Fractional Kelly: A Practical Framework

Here’s how to implement Kelly-based position sizing in practice:

Step 1: Define Your Edge

Identify your investment strategy and calculate your historical win rate, average win size, and average loss size. You need at least 30-50 prior trades to have statistical confidence.

Step 2: Calculate Expected Value

For each new opportunity, estimate the probability of success (p) and the payoff ratio (b). If historical data applies, use it; otherwise, develop scenario-based estimates.

Step 3: Compute Kelly Fraction

Use the formula to calculate the full Kelly percentage.

Step 4: Apply Fractional Kelly

Use 25-50% of the calculated fraction. Most sophisticated investors use quarter-Kelly to half-Kelly.

Step 5: Enforce Position Limits

Set maximum position sizes (e.g., no single position exceeds 10% of portfolio) regardless of Kelly calculations. This protects against correlation surprises and catastrophic single-position failure.

Step 6: Rebalance Regularly

As positions grow or shrink, rebalance back to target allocations. This enforces discipline and prevents position drift.

This framework combines the mathematical rigor of Kelly with practical safeguards for the uncertainty inherent in real investing.

Real-World Examples: How Professionals Use Kelly

Several successful investment approaches incorporate Kelly-style thinking:

Algorithmic Trading Firms

Quantitative hedge funds with high-volume, low-latency trading systems can estimate edge accurately. These firms extensively use Kelly-based sizing for individual bets, which compounds into superior long-term returns. Renaissance Technologies, while secretive, is known to use probability-based position sizing consistent with Kelly principles.

Sports Betting Syndicates

Professional sports bettors developed Kelly methodology to optimal precision. The best syndicates maintain detailed records of their edge on thousands of bets, continuously update probability estimates, and use fractional Kelly sizing to compound wealth reliably.

Value Investors

Warren Buffett uses concentration (a Kelly-style principle) by investing larger amounts in highest-conviction ideas. While he doesn’t explicitly use Kelly formulas, his statements about position sizing (“put large amounts of capital in your highest-conviction ideas”) reflect Kelly-aligned thinking.

Options Traders

Options traders frequently use Kelly-based sizing because options pricing models (like Black-Scholes) explicitly estimate probabilities. A trader with models suggesting a 65% probability of profitability on a call option spread can directly apply Kelly to size that position.

Common Mistakes and How to Avoid Them

Mistake 1: Using Full Kelly Without Estimation Error Margins

Solution: Always use fractional Kelly (25-50%) to create a safety margin against probability estimation errors.

Mistake 2: Calculating Kelly on Insufficient Historical Data

Solution: Require minimum 30-50 trades before considering historical win rates reliable. Until then, use conservative prior probability estimates.

Mistake 3: Ignoring Position Correlation

Solution: Apply maximum position size limits to prevent excessive concentration risk, regardless of individual position Kelly calculations.

Mistake 4: Static Probability Estimates

Solution: Update probability estimates as new information arrives. Use Bayesian methods to incorporate changes in market conditions.

Mistake 5: Not Accounting for Tax Implications

Solution: Adjust Kelly calculations for after-tax payoffs, not pre-tax. In taxable accounts, after-tax expected value is what matters.

Conclusion: The Path to Optimal Position Sizing

The Kelly Criterion represents a sophisticated approach to one of investing’s most critical decisions: how much capital to allocate to each position. By grounding position sizing in mathematical principles rather than intuition or arbitrary rules, Kelly offers a path to superior long-term wealth accumulation.

However, practical application requires wisdom beyond the formula itself. Use fractional Kelly as insurance against probability estimation errors. Enforce position limits to manage unforeseen correlations. Update your estimates as new information arrives. These safeguards transform Kelly from an elegant but fragile theory into a robust practical tool.

Whether you’re a systematic trader with decades of data, a value investor with concentrated conviction, or an options trader with probability models, incorporating Kelly-style thinking into your position sizing can materially improve your long-term returns. The key is matching the sophistication of your approach to the reliability of your probability estimates—using more Kelly precision when you have better data, and defaulting to fractional Kelly and position limits when uncertainty is high.

In the end, optimal position sizing isn’t about finding the magic percentage. It’s about making each dollar work as hard as possible while sleeping soundly knowing that no single position can derail your long-term financial plan.

I think the most underrated aspect here is

Published by

Rational Growth Editorial Team

Evidence-based content creators covering health, psychology, investing, and education. Writing from Seoul, South Korea. View all posts by Rational Growth Editorial Team