Instructors

Brett Bernstein and David Frohardt-Lane

Date, time, location

Blended Lecture: Friday, 1:00PM-3:40PM

Office hours

TBD

Course Description

Professional sports provides the ideal setting for teaching statistics (a perspective shared by the eminent statisticians David Donoho and Andrew Gelman). Students taking this course will learn to build the type of models that have transformed the industry over the past decades. By developing predictive models for sports matches, students will gain a deep intuitive understanding of the statistical tools needed to become successful data scientists.

At its core, this is a course about predicting sports outcomes using statistical models based on historical game data, and information from prediction markets. This will directly prepare students for careers in sports analytics and will be useful for students pursuing careers in quantitative finance and other industries. By studying and building these sports models, students will also learn skills that apply beyond the confines of sports analytics.

Rough Syllabus

Some of the topics we cover may include:

  1. Luck vs. Skill: How distinguishing these factors plays an essential role in modeling
  2. Generalized Linear Models
  3. Soccer Models
  4. Understanding Prediction Markets, and Market-implied Match Probabilities
  5. Sports Betting
  6. Baseball Models
  7. Markov Chain-based Game Simulations
  8. Golf Models, and Spatial Data Analysis
  9. Football Models
  10. Basketball Models

Prerequisites

  1. DS-GA 1001: Introduction to Data Science or something equivalent that covers the core data science topics in Python
  2. DS-GA 1002: Probability and Statistics for Data Science or something equivalent
  3. Course is well-paired with DS-GA 1003 Machine Learning , but this is not a prerequisite