Going Deep: models for continuous-time within-play valuation of game outcomes in american football with tracking data

Continuous-time assessments of game outcomes in sports have become increasingly common in the last decade. In American football, only discrete-time estimates of play value were possible, since the most advanced public football datasets were recorded …

Bayesian Baby Steps: Normal Next Steps

Enter marquis de Laplace In my first post on Bayesian data analysis, I did a brief overview of how Bayesian updating works using grid approximation to arrive at posterior distributions for our parameters of interest, such as a wide receiver’s catch rate. While the grid-based approach is simple and easy to follow, it’s just not practical. Before we turn to MCMC, in this post we’ll cover the popular approach known as Laplace approximation1, aka quadratic approximation.

nflWAR: a reproducible method for offensive player evaluation in football

Unlike other major professional sports, American football lacks comprehensive statistical ratings for player evaluation that are both reproducible and easily interpretable in terms of game outcomes. Existing methods for player evaluation in football …

Bayesian Baby Steps: Intro with JuJu

First steps I was originally thinking of writing a blog post about multilevel models (aka hierachical, mixed, random effects) because of how useful they are for measuring player performance in sports1 (shameless self promotion for nflWAR here!). But the more I thought about it, the more I realized how ill-minded of an idea that was. Instead, I want to build up the intuition for how and why one would want to use a full Bayesian multilevel model.


Along with Maksim Horowitz and Sam Ventura, I have developed the nflscrapR package in R which allows for easy access of publicly available NFL play-by-play data. We provide estimates for the expected points and win probability for every play based on our fully reproducible methodology explained in our paper available for free on ArXiv. The nflscrapR package is frequently used by the football analytics community appearing in articles on The Athletic and FiveThirtyEight, as well as shared by the NFL Director of Data and Analytics.