Application of the Kelly Criterion to Prediction Markets

Bernhard K Meister

2024-12-19

Abstract:Betting markets are gaining in popularity. Mean beliefs generally differ from prices in prediction markets. Logarithmic utility is employed to study the risk and return adjustments to prices. Some consequences are described. A modified payout structure is proposed. A simple asset price model based on flipping biased coins is investigated. It is shown using the Kullback-Leibler divergence how the misjudgment of the bias and the miscalculation of the investment fraction influence the portfolio growth rate.

Portfolio Management

What problem does this paper attempt to address?

This paper attempts to solve two main problems:

1. **Application of the Kelly Criterion in Prediction Markets**:
- The paper explores how to apply the Kelly Criterion to prediction markets in order to study the differences between expected values and market prices. Unlike traditional financial markets, the prices in prediction markets have not only a lower limit but also an upper limit, which changes the market dynamics and investment opportunities.
- The Kelly Criterion uses a logarithmic utility function to derive the optimal investment strategy, maximizing the capital growth rate and the associated logarithmic utility. The paper points out that it is incorrect to directly regard the prediction market prices as probabilities. So, how large is this gap? How can the payout structure of bets be adjusted to reduce this gap?

2. **Simple Asset Price Models within a Finite Time**:
- Within a finite time frame, study simple asset price models based on the biased - coin model. Specifically, analyze the impact of misjudgment of the bias or miscalculation of the investment proportion on the portfolio growth rate.
- Use the Kullback - Leibler divergence to quantify the specific impact of these misjudgments on the portfolio growth rate.

### Specific Content

#### 1. Application of the Kelly Criterion in Prediction Markets

- **Relationship between Market Prices and Probabilities**:
- The prices in prediction markets are not equivalent to the probabilities of events occurring. The paper mentions that market prices are affected by factors such as risk aversion and leverage effects, and thus deviate from the actual probabilities.
- The paper proposes an optimized investment proportion formula. When the market price is \( p \) and the investor's subjective belief is \( q \), the optimal Kelly fraction \( f \) is:
\[
f=\frac{q - p}{(1 - q)(1 - p)}=\frac{Q - P}{1 + Q}
\]
where \( P = \frac{p}{1 - p} \) and \( Q=\frac{q}{1 - q} \)

- **Gap between Market Prices and Average Beliefs**:
- The maximum possible gap between market prices and the average beliefs of investors is studied. Assume that there are only two investors in the market, one is certain that the event will not occur, and the other is certain that the event will occur. The relationship between market prices and average beliefs is obtained through market - clearing conditions.
- The results show that even if the views of investors in the market are extremely polarized, the market price may still be very different from the average belief.

#### 2. Simple Asset Price Models within a Finite Time

- **Performance of Different Strategies in Double - or - Nothing Games**:
- Use the Kullback - Leibler divergence to analyze the performance of different strategies in double - or - nothing games. Specifically, analyze the impact of misjudgment of the random walk bias and misjudgment of the optimal investment proportion on the portfolio growth rate.
- For example, for a small change \( \epsilon \) in probability \( p \), the change in the Kullback - Leibler divergence is:
\[
D\left(\frac{k}{N} \| p+\epsilon\right)-D\left(\frac{k}{N} \| p\right)=\frac{p - k/N}{p(1 - p)}\epsilon+O(\epsilon^{2})
\]

### Conclusion

Information theory, even in its simplest form, can provide profound insights into prediction markets. The article shows how to modify the market by stretching or compressing the payout ratios near the boundaries, thereby reducing the gap between market prices and average beliefs. At the same time, the Kullback - Leibler divergence can be used as a tool to analyze the performance of different strategies in double - or - nothing games, especially in cases of misjudgment of random walk bias and investment proportion.