Chapter 13 Further Reading: Value Betting Theory and Practice
The following annotated bibliography provides resources for deeper exploration of the topics introduced in Chapter 13. Entries are organized by category and chosen for their relevance to probability estimation, value identification, performance evaluation, and adaptive betting strategies.
Books: Value Betting and Probability
1. Silver, Nate. The Signal and the Noise: Why So Many Predictions Fail --- but Some Don't. Penguin Press, 2012. Silver's widely read book on prediction includes a detailed chapter on sports betting that illustrates the challenge of separating signal from noise in model-based probability estimation. His discussion of calibration, overconfidence, and the difficulty of beating efficient markets provides excellent context for Section 13.1's treatment of model versus market probabilities. The chapter on poker is also relevant, as it addresses edge detection and adaptation under uncertainty.
2. Poundstone, William. Fortune's Formula: The Untold Story of the Scientific Betting System That Beat the Casinos and Wall Street. Hill and Wang, 2005. The definitive popular account of the Kelly criterion and its development by John Kelly, Claude Shannon, and Ed Thorp. Poundstone traces the connection between information theory and optimal bet sizing, providing historical context for Section 13.2's Kelly-based value assessment framework. The book's treatment of the tension between full Kelly and fractional Kelly --- and the practical reasons most professionals use a fraction --- is directly relevant to the confidence-adjusted Kelly recommendation in Chapter 13.
3. Mallios, William S. Sports Betting: Fundamentals of Probability and Statistics. Trafford Publishing, 2008. A rigorous statistical treatment of sports betting probability estimation, including detailed sections on sample size requirements, confidence intervals, and the regression-to-the-mean phenomenon. Mallios's analysis of how long it takes to distinguish a skilled bettor from a lucky one directly informs Section 13.4's treatment of the sample size problem.
4. Miller, Ed. The Logic of Sports Betting. The Hendon Mob, 2019. Miller's practical guide includes an excellent chapter on "What Makes a Good Bet?" that covers edge calculation, minimum thresholds, and the relationship between edge and Kelly sizing. His treatment of value identification complements Chapter 13's mathematical framework with practical wisdom about which edges are real and which are illusory.
Academic Papers: Prediction Models and Calibration
5. Glickman, Mark E., and Jones, Albyn C. "Rating the Chess Rating System." Chance, 12(2), 1999, pp. 21-28. A thorough examination of the Elo rating system's mathematical properties, including its limitations and extensions. The authors discuss how the scaling parameter (400 in standard Elo) affects probability calibration and how the K-factor controls the model's responsiveness to new information. Essential background for Chapter 13's Elo model implementation.
6. Hvattum, Lars Magnus, and Arntzen, Halvard. "Using ELO Ratings for Match Result Prediction in Association Football." International Journal of Forecasting, 26(3), 2010, pp. 460-470. Demonstrates the application of Elo ratings to football prediction and compares the resulting probability estimates to closing market odds. The authors find that Elo provides a reasonable baseline but is outperformed by the market. This finding directly supports Chapter 13's recommendation to use Bayesian combination rather than relying on Elo alone.
7. Gneiting, Tilmann, and Raftery, Adrian E. "Strictly Proper Scoring Rules, Prediction, and Estimation." Journal of the American Statistical Association, 102(477), 2007, pp. 359-378. The definitive academic treatment of proper scoring rules, including the Brier score and log loss used in Chapter 13's calibration analysis. The authors prove that the Brier score and logarithmic score are strictly proper --- meaning they incentivize honest probability reporting --- which justifies their use as model evaluation metrics.
8. Platt, John. "Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods." Advances in Large Margin Classifiers, MIT Press, 1999, pp. 61-74. Introduces Platt scaling, a method for calibrating probability outputs from machine learning classifiers. While the chapter focuses on Elo and logistic regression (which produce naturally calibrated outputs), bettors who use more complex models (random forests, neural networks, SVMs) will need calibration techniques like Platt scaling to convert model outputs to usable probabilities.
Academic Papers: Market Efficiency and Edge Detection
9. Kaunitz, Lisandro, Zhong, Shenjun, and Kreiner, Javier. "Beating the Bookies with Their Own Numbers --- and How the Online Sports Betting Market Is Rigged." arXiv preprint, 2017. Demonstrates a systematic approach to identifying value by comparing odds across sportsbooks --- a market-based approach to value betting. The authors achieve positive returns by betting on outcomes where at least one book's odds deviate significantly from the consensus. Their finding that books systematically limit winning accounts is relevant to Chapter 13's discussion of edge decay through account restrictions.
10. Stern, Hal S. "On the Probability of Winning a Football Game." The American Statistician, 45(3), 1991, pp. 179-183. An early application of statistical modeling to football outcome prediction. Stern's model uses a normal distribution for point spreads and demonstrates that even simple models can produce reasonably calibrated probability estimates. This paper provides the historical foundation for the model-based approaches in Section 13.1.
11. Forrest, David, Goddard, John, and Simmons, Robert. "Odds-Setters as Forecasters: The Case of English Football." International Journal of Forecasting, 21(3), 2005, pp. 551-564. Examines the forecasting accuracy of bookmaker odds versus statistical models in English football. The authors find that bookmaker odds are generally well-calibrated but that models can add incremental value, particularly for less popular matches. This supports Chapter 13's Bayesian combination approach: the market is good but not perfect, and models can contribute marginal information.
12. Borghesi, Richard. "Widespread Corruption in Sports Gambling: Fact or Fiction?" Southern Economic Journal, 74(4), 2008, pp. 1063-1074. Investigates whether closing line movements contain genuine information about game outcomes. The finding that closing lines are systematically more accurate than opening lines validates CLV as a skill metric and supports the use of closing probabilities as the "true probability" benchmark in Chapter 13's framework.
Websites and Data Sources
13. FiveThirtyEight Elo Ratings (projects.fivethirtyeight.com) FiveThirtyEight's publicly available Elo ratings for NFL, NBA, and other sports provide both a data source and a calibration benchmark. Their Elo methodology, including sport-specific K-factors, home-field adjustments, and margin-of-victory multipliers, closely mirrors the model described in Section 13.1. Comparing your model's outputs to FiveThirtyEight's is a useful sanity check.
14. Pinnacle Sports Blog (pinnacle.com/betting-resources) Pinnacle's educational content includes rigorous articles on value betting, probability estimation, and the relationship between edge and Kelly sizing. Their article "Expected Value Explained" provides an accessible introduction to the concepts formalized in Chapter 13. Pinnacle's closing lines are widely used as the "market probability" benchmark in CLV calculations.
15. Pro Football Reference (pro-football-reference.com) The primary free data source for NFL historical statistics, including game results, team statistics, and advanced metrics. Pro Football Reference's data can be used to build and backtest the Elo and feature-rich models described in Section 13.1. The "Game Finder" tool is particularly useful for filtering games by rest, weather, and other situational factors.
16. Basketball Reference (basketball-reference.com) The NBA equivalent of Pro Football Reference. Provides pace, efficiency, and advanced box score statistics needed for the NBA-focused models discussed in the case studies. The "Team Per Game" and "Team Advanced" pages provide the efficiency metrics used as features in Section 13.1's logistic regression example.
17. Unabated (unabated.com) A professional-grade betting analytics platform that provides real-time odds comparison, expected value calculations, and CLV tracking. Unabated's "EV Calculator" implements the value assessment framework described in Section 13.2, making it a useful commercial alternative for bettors who prefer not to build their own tools.
Podcasts and Media
18. Circles Off Podcast (hosted by Rufus Peabody and Adam Chernoff) Sharp bettors who frequently discuss value identification, model building, and the practical challenges of maintaining an edge over time. Episodes covering model calibration, CLV tracking, and the importance of process over outcomes are directly relevant to Chapter 13. Peabody's discussion of how he evaluates whether his models still have edge provides real-world context for Section 13.5's edge decay framework.
19. The Power Rank Blog and Podcast (thepowerrank.com, hosted by Ed Feng) Ed Feng, a former physics professor, applies quantitative methods to sports prediction. His blog and podcast cover Elo models, logistic regression, and the practical challenges of building sports prediction systems. His transparent discussion of model performance (including periods where his models underperform the market) illustrates the honest self-assessment that Chapter 13 advocates.
20. Pinnacle Podcast (pinnacle.com) Pinnacle's podcast features interviews with sharp bettors, quantitative analysts, and market makers. Episodes on "How We Set Lines" and "What Makes a Sharp Bettor" provide insider perspective on the adversarial dynamics between sportsbooks and bettors --- the dynamics that drive the edge lifecycle described in Section 13.5.
How to Use This Reading List
For readers working through this textbook sequentially, the following prioritization is suggested:
- Start with: Miller (entry 4) for practical value betting guidance, and Pinnacle's blog (entry 14) for expected value fundamentals.
- Go deeper on probability models: Silver (entry 1) for intuition, Glickman and Jones (entry 5) for Elo theory, and FiveThirtyEight (entry 13) for calibration benchmarks.
- Go deeper on calibration and scoring: Gneiting and Raftery (entry 7) for the mathematical foundations, and Hvattum and Arntzen (entry 6) for applied comparison of models to markets.
- Go deeper on edge detection and adaptation: Kaunitz et al. (entry 9) for market-based value identification, and the Circles Off podcast (entry 18) for practical edge maintenance strategies.
- For building your own models: Pro Football Reference (entry 15) or Basketball Reference (entry 16) for data, and the Python ecosystem described in Chapter 12's further reading for implementation tools.
- For Kelly criterion depth: Poundstone (entry 2) for history and intuition, and Mallios (entry 3) for the statistical foundations of bet sizing under uncertainty.
Many of these resources are also referenced in Chapters 12 and 14, reflecting the close interconnection between line shopping, value identification, and bankroll management.