This month we welcome Austin Rochford to the BDA meetup. Austin is the Chief Data Scientist at Monetate. He is a PyMC3 developer, a recovering mathematician and is passionate about math education, Bayesian statistics, and machine learning. His writing is available online at http://austinrochford.com.
At least since the publication of Moneyball in 2003 (a realistically much earlier), advanced sports statistics have become increasingly mainstream. Often, the developers of these advanced statistics have been passionate fans with a quantitative bent who are not formally trained in statistics. This situation has led to important discoveries, but often the quantification of uncertainty in these advanced statistics is based on ad-hoc rules-of-thumb.
This talk will demonstrate how Bayesian hierarchical models and probabilistic programming provide a (relatively) user-friendly approach to quantifying uncertainty in sports analytics through examples from the MLB and NHL.