Social Networks in the NBA

Introduction

As a huge fan of basketball I was curious to see if if the concepts taught in CSCC46 could be applied to a network analysis of the NBA. Turns out a student at UCLA named Biancheng Wang wrote a thesis analyzing social networks in the NBA. In this paper they analyze the network of the Instagram following for the NBA All-Stars for the previous five seasons. This analysis covers clustering and density, variables that may cause edges between players and if any “gangs” have formed between All-Stars.

Data

The dataset that Biancheng researched were the NBA All-Stars from the 2014-2019 seasons. Few players were exempt from the analysis due to their less prominent presence on social media or their retirement from the NBA. In total 44 All-Stars were part of the analysis and placed into two networks. One network represented the instagram following of the players, a directed graph where each node is a player and each edge represents a relationship where A follows B on instagram (if the edge goes from A->B). The second network that was part of this thesis was a graph of teammate relationships, where nodes are NBA players and edges represent if the teammates had ever played together in an undirected network.

Since instagram follows do not represent real friendships the author decided to break down ties between players into 4 categories that could be contributing factors to their relationship. Within these 4 categories 22 node attributes were considered as well.

It is also worth noting that the student used a random graph model to further strengthen their analysis. It is called the Exponential-family Random Graph Model (ERGM) and here are some details about its definition.

Results

Here are images of the networks described in the introduction:

The density of the teammate network (edges/# of possible edges) is 0.077 where as the density of the Instagram following network is 0.322. The average length of the shortest path between 2 players in the Instagram network is 1.747 showing that there is a lot of clustering going on in the second network. The author found that team and nationality homophily seriously affected these relationships, players within the same team or nationality generally were tied together. Also, players who played for the same state in University had strong homophily between each other. The study found strong clustering between American black players versus international white players, another example of homophily. This effect then ties into the basketball factor as it is much more likely for two players to be tied together with more experience in the NBA. Business factors did not play a large role in the formation of these relationships, having the same agent or brand really did not play a role in the network structure.

Conclusion

The results presented by this study were generally what you would expect, a few things were surprising, such as the lack of clustering between white international NBA players. I really enjoyed reading this technical yet elegant paper on a league that I am so involved in, if you would like to read the article in more depth the citation can be found below!

Citation

Wang, Biancheng, and Mark S. Handcock. “A Social Network Analysis of NBA Players.” ProQuest Dissertations and Theses, University of California, Los Angeles, 2019, p. 48. ProQuest Dissertations & Theses Global, http://myaccess.library.utoronto.ca/login?qurl=https%3A%2F%2Fwww.proquest.com%2Fdissertations-theses%2Fsocial-network-analysis-nba-players%2Fdocview%2F2299502925%2Fse-2%3Faccountid%3D14771. Accessed 22 Oct. 2022.