Predictive Modeling and Statistical Analysis in Sports

In order to fully examine the breadth of possibility in sports data analysis, we look at two areas of differing scope. The first is scalable and generalizable, exploring the expected performance of athletes in track and field events by means of a datacentric model. It takes Athletic.net data using an in-house web scraper and fits past results against a corpus in order to predict future results in the same race distance. Furthermore, it builds a recommendation system based on composite athlete profiles and displays similar athletes. The second is more specific, examining the effect of momentum in professional NBA basketball with respect to both players and teams. Momentum is defined via a set of enumerable conditions and its model is created by means of a variation of the same inhouse web scraper. Together, these experimental analyses are represented in the same user interface. As such, they demonstrate the usefulness and applicability of predictive modeling across many subfields within sportsdata science.