Entropy-Based Strategies for Multi-Bracket Pools.

Area of Science:

• Decision Science
• Probability Theory
• Computational Statistics

Background:

• Parimutuel betting literature often focuses on single event probabilities or optimal wagering strategies.
• Existing methods become intractable for multi-event betting pools (e.g., March Madness brackets) with higher dimensions.
• Predicting tuples of events requires new approaches beyond traditional single-event strategies.

Purpose of the Study:

• To address the "multi-brackets problem": tractably generating optimal predicted tuples for multi-event betting pools.
• To develop a scalable and efficient method for generating multiple predictions for complex event tuples.
• To explore entropy-based distributions for generating these predictions.

Main Methods:

• Formulated the "multi-brackets problem" for predicting tuples of events.
• Developed an entropy-based approach to generate *n* independent predicted tuples.
• Focused on a simpler setting by using a distribution with optimal entropy.

Main Results:

• The proposed entropy-based approach is tractable and scalable.
• The method demonstrates good performance in generating predicted tuples.
• Successfully addresses limitations of traditional methods in higher-dimensional betting scenarios.

Conclusions:

• The entropy-based approach provides an effective solution for the multi-brackets problem.
• This method offers a scalable and computationally feasible way to handle complex betting pools.
• The findings advance strategies for predicting outcomes in multi-event scenarios.