The Kelly Criterion and Logarithmic Utility Explained
This article explores the Kelly Criterion, a mathematical formula used to determine the optimal size for a series of bets or investments to maximize long-term wealth. It breaks down the mechanics of the formula, the critical role of logarithmic utility in balancing risk and reward, and how maximizing the expected geometric growth rate ensures maximum compounding returns while mathematically eliminating the risk of total ruin.
What is the Kelly Criterion?
Developed by John L. Kelly Jr. in 1956 while working at Bell Labs, the Kelly Criterion is a formula that calculates the exact proportion of your total capital (bankroll) you should wager on a favorable bet.
For simple bets with two outcomes (win or lose), the formula is:
f* = (bp - q) / b
- f* is the fraction of the current bankroll to wager.
- b is the net odds received on the wager (e.g., betting $1 to win $2 means b = 2).
- p is the probability of winning.
- q is the probability of losing (which is 1 - p).
For example, if you have a 60% chance of winning (p = 0.60, q = 0.40) on an even-money bet (b = 1), the formula dictates: f* = ((1 * 0.60) - 0.40) / 1 = 0.20. To maximize long-term growth, you should bet 20% of your bankroll.
The Flaw of Expected Value
To understand why the Kelly Criterion works, you must first understand why standard expected value fails in repeated bets.
If you are offered a bet with a 60% chance of doubling your money and a 40% chance of losing it all, a strategy based purely on maximizing expected value dictates that you should bet 100% of your wealth every time. However, if you bet 100% of your wealth repeatedly, you are mathematically guaranteed to eventually hit that 40% loss and go bankrupt.
Standard expected value optimizes the arithmetic mean (the average outcome of a single bet played by an infinite number of people). However, an individual investor experiences wealth sequentially through compounding. Compounding is multiplicative, meaning it is governed by the geometric mean.
How Logarithmic Utility Solves the Problem
Instead of maximizing expected wealth, the Kelly Criterion maximizes the expected logarithm of wealth. In economics and mathematics, this is known as using a logarithmic utility function: U(W) = ln(W), where W is your wealth.
Using logarithms achieves three critical mathematical objectives for long-term growth:
1. Converting Multiplication into Addition
Wealth compounds multiplicatively. If you win 50% and then lose 50%, your wealth is multiplied by 1.5 and then by 0.5, leaving you with 0.75 (a 25% loss). Logarithms convert multiplication into addition: ln(a * b) = ln(a) + ln(b). By taking the logarithm of wealth, the Kelly Criterion transforms a complex compounding problem into a simpler additive one, allowing you to find the highest geometric growth rate.
2. Accounting for Diminishing Marginal Utility
The logarithmic curve is concave, meaning it flattens out as numbers get larger. In practical terms, this reflects the reality that losing half your bankroll hurts much more than winning half your bankroll helps. A 50% loss requires a 100% gain just to break even. Logarithmic utility naturally penalizes volatility and large drawdowns, prioritizing survival over reckless gains.
3. Preventing Total Ruin
In mathematics, the logarithm of zero is negative infinity: ln(0) = -∞. Because the Kelly Criterion seeks to maximize the expected logarithm of wealth, any strategy that carries even a tiny risk of losing 100% of your bankroll results in an expected utility of negative infinity. Therefore, the logarithmic function acts as an absolute mathematical barrier against bankruptcy. You will never be told to bet 100% of your bankroll unless the probability of winning is explicitly 100%.
The Mathematical Derivation
To find the optimal bet size using logarithmic utility, we look at the expected growth rate (G) of a bankroll after one bet:
G = p * ln(1 + bf) + q * ln(1 - f)
- 1 + bf represents your new wealth if you win.
- 1 - f represents your new wealth if you lose.
If you plot this equation on a graph where the x-axis is the bet size (f) and the y-axis is the growth rate (G), it forms an upside-down parabola.
- Under-betting (to the left of the peak) results in lower risk, but sub-optimal growth.
- Optimal betting (the exact peak of the curve) provides the maximum possible long-term growth rate.
- Over-betting (to the right of the peak) increases risk and decreases your growth rate due to the mathematical drag of large losses. Bet too far to the right, and your growth rate turns negative, guaranteeing eventual ruin despite having a statistical edge.
Using calculus, we find the exact peak of this curve by taking the derivative of the growth rate equation with respect to f, setting it to zero, and solving for f. The result of that calculus is the Kelly Criterion formula: f* = (bp - q) / b.