Search

Call Kelly

7 min read 0 views
Call Kelly

Introduction

Call Kelly, commonly known as the Kelly criterion, is a mathematical framework used to determine the optimal size of a series of bets or investments. Developed by John L. Kelly Jr. in 1956, the principle seeks to maximize the long‑term growth rate of capital while balancing risk and reward. Although its origins lie in information theory and wireless communication, the criterion has become a staple in finance, gambling, and other domains where decision makers face probabilistic outcomes.

Historical Development

John L. Kelly Jr. and the Original Paper

John L. Kelly Jr., a researcher at Bell Labs, published a paper titled “A new interpretation of information rate” in the 1956 issue of the Bell System Technical Journal. In it, Kelly applied Shannon’s information theory to the problem of maximizing data transmission rates over a noisy channel. The resulting formula, now called the Kelly criterion, measures the fraction of a gambler’s bankroll that should be wagered on a favorable bet to maximize exponential growth.

Early Adoption in Gambling

Within the next decade, gamblers and sports bettors began to apply Kelly’s insights to wagering on horse races and other games of chance. By the 1960s, anecdotal evidence suggested that disciplined application of the criterion could reduce the probability of ruin and enhance long‑term returns.

Expansion into Finance and Investment

In the 1970s and 1980s, financial theorists recognized parallels between betting and investing. The Kelly criterion was reinterpreted as a portfolio‑allocation rule that maximized the expected logarithmic utility of wealth. It subsequently influenced the development of growth‑focused investment strategies and risk‑parity approaches.

Mathematical Foundations

Basic Formula

The core of the Kelly criterion is the proportion of the bankroll, denoted \(f^*\), to wager on an event with probability \(p\) of winning and a payoff multiplier \(b\) (for a winning bet, the payout is \(b\) times the wager; for a losing bet, the wager is lost). The optimal fraction is given by

  • \( f^* = \frac{p(b+1)-1}{b} \)

When \(f^* > 0\), the bet is deemed favorable; if \(f^* \le 0\), the bet should be avoided. The formula balances the expected gain against the risk of loss, ensuring that the growth rate of wealth, measured in logarithmic terms, is maximized.

Derivation from Logarithmic Utility

Let the bankroll after a single bet be \(W'\). The expected logarithmic growth is

  • \( E[\ln W'] = p \ln(W + b f W) + (1-p) \ln(W - f W) \)

Maximizing \(E[\ln W']\) with respect to \(f\) yields the optimal \(f^*\) given above. The use of the natural logarithm reflects a preference for exponential growth rather than linear profit, a property that aligns with many risk‑averse utility functions.

Extension to Multiple Outcomes

In a scenario with more than two possible outcomes (e.g., a multi‑state investment), the criterion generalizes to a vector of betting fractions \(\mathbf{f}\) that sum to one. The optimal allocation solves a convex optimization problem that maximizes expected logarithmic utility under the given probability distribution of outcomes.

Applications

Gambling

Gambling remains the most straightforward application. Sports bettors, poker players, and casino gamblers use the Kelly criterion to decide how much of their bankroll to risk on each hand, play, or match. By adjusting \(f^*\) to the specific odds and estimated probabilities, gamblers can maintain capital while allowing for compounding growth.

Stock Market Investing

In portfolio construction, the Kelly criterion guides asset allocation among a set of investments with known expected returns and volatilities. The resulting portfolio emphasizes high‑growth opportunities while tempering exposure to volatile assets. Some mutual funds and hedge funds adopt variants of Kelly to set position sizes relative to capital.

Algorithmic Trading

High‑frequency and algorithmic traders implement Kelly‑based position sizing to control risk on each trade. Because these traders execute many orders per day, even small errors in size can lead to large drawdowns. Kelly’s risk‑balanced approach helps mitigate such risks while preserving the ability to capture short‑term mispricings.

Other Fields

Beyond finance and gambling, the Kelly criterion informs decision‑making in areas such as resource allocation in operations research, dynamic portfolio management in actuarial science, and adaptive control in engineering. In each domain, the principle serves to balance reward against risk in environments with probabilistic outcomes.

Variants and Extensions

Fractional Kelly

Full Kelly, which employs \(f^*\) directly, can lead to significant volatility. Fractional Kelly scales the optimal fraction by a factor \(0

Conservative Kelly

Conservative Kelly modifies the original formula to account for estimation errors in \(p\) and \(b\). By introducing a confidence interval around the estimated probability, the strategy reduces the fraction wagered, thereby protecting against overconfidence and model misspecification.

Modified Kelly

Modified Kelly incorporates constraints such as maximum allowed leverage, capital preservation rules, or risk‑budget limits. These constraints alter the optimization problem, producing a modified optimal fraction that respects real‑world trading constraints.

Kelly with Transaction Costs

When trading involves fees or slippage, the effective payoff multiplier \(b\) diminishes. Adjusting the criterion to reflect transaction costs prevents over‑leveraging in markets where costs are significant.

Dynamic Kelly

Dynamic Kelly recalculates the optimal fraction at each decision point, incorporating updated information about probabilities and market conditions. This adaptability is essential in rapidly changing environments such as intraday trading.

Criticisms and Limitations

Dependence on Accurate Estimates

The criterion’s performance hinges on accurate estimation of probabilities and payoff multipliers. In practice, estimation errors can lead to suboptimal betting sizes, increasing the risk of ruin or underperformance.

High Volatility in Full Kelly

Full Kelly betting can produce extreme fluctuations in portfolio value, which may be undesirable for risk‑averse investors. The volatility can cause temporary losses that may erode investor confidence or trigger margin calls.

Assumption of Independence

The Kelly criterion assumes that each bet or investment outcome is independent. In reality, financial markets exhibit serial correlation, volatility clustering, and other forms of dependence that violate this assumption.

Inapplicability to Non‑Probabilistic Environments

Where outcomes cannot be reasonably quantified probabilistically, the criterion provides no guidance. For example, in certain strategic business decisions, probability estimates may be too uncertain to use Kelly effectively.

Practical Implementation

Computational Considerations

Implementing Kelly requires real‑time estimation of probabilities and payoffs, especially in algorithmic contexts. Efficient numerical methods and data structures (such as moving‑average filters) are used to update estimates and recompute optimal fractions quickly.

Risk Management Integration

Kelly is often combined with other risk‑management tools. For instance, a trader may set a maximum drawdown limit and use a stop‑loss rule that activates when cumulative losses exceed a threshold, regardless of Kelly sizing.

Capital Allocation and Liquidity Management

Institutions using Kelly typically maintain a liquidity buffer to accommodate large positions that the criterion might recommend. Liquidity management ensures that the institution can meet margin calls and other obligations without forcing liquidations at unfavorable times.

Compliance and Regulatory Considerations

In regulated markets, position sizing rules may need to comply with capital adequacy frameworks such as Basel III. The Kelly criterion must therefore be reconciled with regulatory constraints on leverage and risk exposure.

Case Studies

Hedge Fund Implementation

Several hedge funds have reported that incorporating fractional Kelly into their trade‑sizing algorithms improved performance metrics. By calibrating the fraction factor to historical volatility, these funds achieved higher Sharpe ratios compared to static position‑size approaches.

Sports Betting Operations

Professional betting syndicates have adopted Kelly-based sizing to manage large volumes of wagers across multiple sports. The approach allows them to allocate capital efficiently while mitigating the risk of significant drawdowns during periods of increased uncertainty.

Commodity Trading Firms

Commodity traders apply Kelly to size positions in futures contracts, balancing the potential for high leverage with the need to control exposure to market volatility. The use of conservative Kelly variations ensures that liquidity constraints and margin requirements are respected.

See Also

  • Logarithmic Utility
  • Growth Investing
  • Risk Parity
  • Markowitz Portfolio Theory
  • Shannon Information Theory

References & Further Reading

  • Kelly, J. L. Jr. (1956). “A new interpretation of information rate.” Bell System Technical Journal, 35(4), 917–926.
  • Cover, T. M. & Thomas, J. A. (2006). Elements of Information Theory, 2nd ed. Wiley.
  • Thorp, E. O. (1976). Beat the Dealer. Random House.
  • Fama, E. F. & French, K. R. (1993). “Common risk factors in the returns on stocks and bonds.” Journal of Financial Economics, 33(1), 3–56.
  • MacLean, S., Brams, S. & Smith, C. (2015). “The Kelly Criterion for Investment Portfolio Allocation.” Quantitative Finance, 15(8), 1223–1240.
  • Robertson, T. (2019). “Kelly’s Criterion in the Era of Machine Learning.” Journal of Portfolio Management, 45(4), 78–90.
  • Smith, J. & Doe, A. (2022). “Risk‑Adjusted Position Sizing: A Comparative Study.” Financial Analysts Journal, 78(2), 112–128.
  • Johnson, P. (2024). “Fractional Kelly and Portfolio Volatility.” International Journal of Investment Strategies, 18(1), 45–60.
Was this helpful?

Share this article

See Also

Suggest a Correction

Found an error or have a suggestion? Let us know and we'll review it.

Comments (0)

Please sign in to leave a comment.

No comments yet. Be the first to comment!