Data-Driven Approaches to NBA Team Evaluation and Building

Abstract

In the National Basketball Association (NBA), it has historically been difficult to build and sustain a team that can consistently compete for championships. Given this challenge, we have developed a series of analyses to support NBA teams in making data-driven decisions. Relying on a variety of datasets, we examined several facets related to the construction of NBA rosters and their performance. In our analysis of on-court performance, we have used clustering algorithms to classify teams in terms of play style, and determined which play styles tend to lead to success. In our analysis of roster construction and transactions, we have investigated the relative value of draft picks and the impact of trades involving draft picks, as well as the effect of roster continuity (i.e. maintaining the same players across seasons) on team success. Additionally, we have developed a model for predicting player contract values and performance versus contract value, which will help teams in identifying the most cost-effective players to acquire. Ultimately, this assembly of analyses, in conjunction, can be used to inform any NBA team’s decisions in its pursuit of success.