Network Science Based Basketball Analytics Paper
Paper I wrote alongside two Graduate students while in my network Science class (CSE 5245) at Ohio State. While the paper goes into full detail on what this project,
here I'll describe a "Cliff Notes" version of the paper. In essence the project looked into a new way of analyzing basketball teams, in a more team-oriented
analysis rather than based on individual performance. Measures such as entropy (how "random" a team is), degree centrality (how much the ball flows through each player on offense),
and uphill-downhill flux (the tendency of a team to get the ball into their best shooter's hands) were used. Results taken from OSU games were compared to a baseline dataset (NBA data) and used to derive conclusions about this Ohio State team.
Since moment-by-moment data exists for NBA data but not college data, this necessitated that we use NBA data, even if it's not necessarily comparable comparable to NCAA data. In order to get data from Ohio State's games, full games were found online for 3 different games in the 2015-2016 season; Kentucky, Iowa (which were wins) and Michigan State (which was a loss). All of the passes during these games were kept track of, and the resulting data was used in a python program to calculate the appropriate measures. More games would have liked to have been evaluated, however due to the amount of time needed to manually type out all the passes, and time constraints, this was not feasible.