In Class 4 we get to what many consider the heart of sports analytics: sports statistics. Statistics have long been a part of sports, but recent times have seen an explosion of new, improved, advanced statistics. In this class we talk about several examples of advanced statistics and then discuss methods for developing new player performance metrics.
Our opening example is the NFL Passer Rating statistic.
This is one of my favorite statistics because it illustrates the fundamentals of this process and includes some thought provoking (crazy?) elements. We break the statistic down and talk about what each element represents. Where things get shaky is when we talk about how the different elements are combined.
How are advance statistics created? There are several approaches, but the most common approach is a Multi Attribute Decision Model. These models involve weighted combinations of multiple sports statistics. To illustrate how this approach works we discuss the On Base Plus Slugging statistic from baseball.
OBPS is a great statistic. It captures key elements of player performance and it has a simple structure. But when we put it under the microscope, we quickly get to the issue of whether the “weights” are correct.
We also spend some time on the Gold Standard method where sports statistics are developed using statistical models to investigate the relationship between some measure of ultimate performance and player performance metrics.
In Class 5, we continue our discussion of sports analytics, but we shift focus to in-game decision making. This discussion will examine classic sports decisions such as whether to swing away or go for the sacrifice bunt. These kinds of decisions exist across all sports. For preparation, I encourage everyone to think about their favorite in-game decision. Examples include pulling the goalie or when to go for 2. Think about the logic of the decision for the immediate situation AND how the decision impacts the how the game plays out AFTER the decision. For the more mathematically inclined there will be some discussion of Markov decision models.
Listen to today's full class here: