Calculated Bets: Computers, Gambling, and Mathematical Modeling to Win

By Steven Skiena

Quick Summary

Computer scientist Steven Skiena documents his attempt to build a mathematical model to beat the pari-mutuel betting markets at jai-alai, a fast-paced Basque sport popular in Florida. The book serves as both a practical case study in quantitative modeling and a broader meditation on the scientific process of building, testing, and refining predictive models under conditions of uncertainty -- skills directly applicable to financial market modeling.

Detailed Summary

The Jai-Alai Problem

Jai-alai is a sport where players compete in a round-robin format, and the betting system uses pari-mutuel pools (win, place, show, exacta, trifecta). Skiena identifies this as an ideal testing ground for quantitative modeling because: the outcomes are objectively measurable; the number of possible outcomes is finite and calculable; the pari-mutuel system means you are competing against other bettors rather than a house edge; and historical data is available for model validation.

Model Building Process

The book walks through the complete scientific process of building a predictive model: gathering and cleaning historical data (from the World Wide Web and other sources); formulating hypotheses about what factors predict outcomes; building mathematical models (including probability distributions, trigonometric functions, cellular automata, and Voronoi diagrams); validating models against held-out data; and iterating based on results. The treatment of unclaimed tickets, withholding taxes, and the actual mechanics of pari-mutuel payoffs adds practical realism.

Key Contributors and Insights

The work of Dario Vlah in developing automated data scraping and analysis tools is highlighted. The analysis of post position effects on outcomes, the distribution of trifecta payoffs, variance analysis, and the identification of non-obvious betting opportunities (boxes, wheeling strategies) demonstrate how quantitative thinking can uncover edges invisible to casual observers.

Broader Lessons for Financial Modeling

While the domain is sports betting, the lessons are directly transferable to financial markets: the importance of validation (testing models against data not used in construction); the danger of overfitting; the distinction between statistical significance and practical profitability (after transaction costs and taxes); the role of variance in determining whether a positive-expectation strategy can survive long enough to realize its edge; and the humbling gap between theoretical models and real-world performance.

Categories

• Quantitative Analysis
• Trading Systems
• Probability & Statistics

Key Takeaways

• The scientific process of model building -- hypothesis, testing, validation, iteration -- is applicable to any domain of uncertainty
• Quantitative models can identify non-obvious edges in betting and financial markets
• Validation against held-out data is essential to avoid overfitting
• The gap between theoretical edge and practical profitability (after costs, variance, and taxes) is substantial
• The skills of data gathering, cleaning, analysis, and model construction are transferable from gambling to financial markets