How We Use Kalman Filters For NBA Player Ratings
What is a Kalman Filter? Why is it used to measure player ratings in the NBA at DraftKings? Let’s dive right in.
To start, A Kalman Filter is an algorithm that allows you to measure the state of an object at a given point in time. Due to the high-score nature of basketball games and the impact an individual player can have on the team and NBA championships, certain players can be worth many points when creating player ratings.
There are several ways to build player ratings, and this information already exists publicly. For example, ESPN has a widely available rating system. These systems aim to capture the impact of each player, both on offense and defense, consolidated into one number. It’s unclear how the player ratings are produced, but we aim to recreate something similar using Kalman Filters.
To put it simply, Kalman Filters takes the current state of an object (e.g., our prediction of the ability of a player), an uncertainty value (e.g., our confidence in that rating of the player's ability and impact — think of a rookie compared with a well-seasoned veteran) and a measurement (e.g., the outcome of a possession) as inputs. We use the algorithm to measure the best estimate of the new value.
There’s much more to dive into about Kalman Filters and NBA player ratings. If you want to take a deep dive into the algorithm, check out our Engineering blog article, where we cover how it compares to other rating systems, an NBA example, and use cases.
To learn more about the Engineering team, check out the work we do.
Check out the latest DraftKings blog posts and meet our global team!Read our Blog
Picture Yourself Here