Kelly (1956): The Kelly Criterion and the Mathematical Limits of Leverage

#The Origin of the Formula

Title: A New Interpretation of Information Rate Author: John L. Kelly, Jr. (Bell Labs) Published: Bell System Technical Journal, 1956

John Kelly, an associate of Claude Shannon at Bell Labs, originally developed his criterion not for Wall Street, but to analyze the transmission of information over noisy communication lines. It was later adapted to describe the optimal bet size for a gambler who possesses an "edge" (inside information) on a horse race.

The equation is profoundly simple and dictates exactly what percentage of your total wealth you should allocate to any given investment.

#The Kelly Equation

For a basic binary outcome (like a coin flip or a simple directional trade), the formula is:

Kelly % = W - [(1 - W) / R]

Where: - W = Winning probability (e.g., historical win rate of the strategy) - R = Win/Loss Ratio (Average Profit per winning trade / Average Loss per losing trade)

The Mathematics of Ruin If your strategy has a 60% win rate and a 1:1 payout ratio, the Kelly Criterion dictates you should allocate exactly 20% of your bankroll to the trade.

The critical insight of Kelly's paper is the concept of Geometric Drag. If you consistently bet more than the Kelly optimal percentage (e.g., betting 40% when the optimal is 20%), you will experience horrifying drawdowns, and mathematically, your long-term compound growth rate will fall to zero.

Over-betting guarantees mathematical extinction.

#Institutional Reality: Fractional Kelly

If the Kelly Criterion maximizes growth rates, why doesn't every hedge fund use it exactly?

Because the Kelly Criterion generates extreme volatility. A full Kelly portfolio will frequently experience 50% to 80% peak-to-trough drawdowns on its way to generating maximum long-term wealth. Human beings (and institutional clients) simply cannot tolerate the psychological pain of a 70% drawdown.

Furthermore, Kelly assumes you know the exact probabilities of the true win rate (W) and risk-reward (R) of the stock market. You do not. The market is not a casino game; statistical edges change.

To compensate for estimation uncertainty and reduce volatility, modern institutional quants almost universally operate on "Half-Kelly" or "Quarter-Kelly" sizing. Betting half of the Kelly optimal amount reduces your theoretical long-term growth rate by roughly 25%, but it dramatically reduces your portfolio volatility by 50% and virtually eliminates the risk of total ruin.