This Week in Sports Analytics: March 01, 2019

1 minute read


This Week in Sports Analytics:’s weekly roundup is here.


A. Solomon Kurz walks through the classic Stein’s Paradox paper and how partial pooling of multilevel models help create better batting average estimates.


Nathan Sandholtz and Jacob Mortensen introduce lineup points lost and lineup point lost contributions to measure efficiency in shot allocation in their Sloan paper titled “Chuckers: Measuring Lineup Shot Distribution Optimality Using Spatial Allocative Efficiency Models”.


The NFL Big Data Bowl concluded on Wednesday and there was a lot of great work. I was particularly interested in the Expected Hypothetical Catch Probability model that Katherine Evans and Sameer Deshpande put together. Katherine has put together an explanation of the work in part 1 of a planned 4 part series.


Prashanth Iyer put together an RShiny App to visualize RAPM data from Evolving Hockey. While baseball has had WAR for over a decade, it is fun to see other sports work on developing their own public versions.


Javier Fernandez, Luke Bornn, and Dan Cervone “Decompose the Immeasurable Sport” by constructing a model that quantifies the expected value of any soccer possession at a frame-by-frame level. Luke Bornn and his lab have consistently put out great work at the Sloan Analytics Conference and he put together a twitter thread on all 11 papers over the last 5 years.


Richard McElreath’s Statistical Rethinking is a must read for anyone looking to learn Bayesian statistics. He recently finished posting his 2019 course lectures here.