Prediction Models with Sports Data in University of Michigan

Week 1

This module introduces the regression models in dealing with the categorical outcome variables in sport contest (i.e., Win, Draw, Lose). It explains the Linear Probability Model (LPM) in terms of its theoretical foundations, computational applications, and empirical limitations. Then the module introduces and demonstrates the Logistic Regression as a better substitute of LPM for the categorical dependent variables.

Week 2

This module explores the relationship between probability and betting markets. It explains the concept of odds, and the relationship between betting odds and probabilities. It then develops a measure of the accuracy of betting odds using sports examples, and assesses the meaning of efficiency in betting markets.

Week 3

This module shows how to forecast the outcome of EPL soccer games using an ordered logit model and publicly available information. It assesses the accuracy of these forecasts against the betting odds and shows that they are remarkably accurate.

Week 4

This module assesses the efficacy of the EPL forecasting model covered in the previous week by replicating the model in the context of three North American team sports leagues (i.e., NHL, NBA, MLB). Specifically, this module shows how to forecast the outcome of NHL, NBA, MLB regular season games using an ordered logit model and publicly available information. It assesses the accuracy of these forecasts against the betting odds.

Week 5

In this module we examine the historical and social consequences of gambling, and the relationship between gambling and statistics. Gambling is explored from the perspective of different ethical and religious systems. Issues of problem gambling are explored and assessed.