This summer I was a paid intern in the JMP Product Testing Group at SAS World Headquarters in Cary, NC. A personal project that I completed this summer through the use of JMP was a poster presented at the SIS Speed Session/Poster Competition at JSM 2014 in Boston. This project was completed under the guidance of Professor Justin Post at NCSU in the Sports Statistics Club.
Abstract: After observing a substantial increase in three point shot attempts in recent NBA seasons, one may wonder what effect this has on how efficient offenses have become. Using spatial NBA shooting data from the 1996-2013 seasons, we investigate trends in how many shots are taken from specific zones of a basketball court and what proportion of these shots are made. Once trends are identified, K-means clustering will be used to identify teams with a distribution of shot attempts that match the trends found, or the "trendsetter teams". To alleviate issues with data accuracy, we attempt to make a smoothed probit regression model of the probability of shot success using Markov Chain Monte Carlo methods. Smoothed models for each of the clusters defined will be compared to determine differences in shooting ability for each of the clusters. To specifically model shot success probability, we use multiple linear regression to determine the importance of shooting area and NBA season on shot success. Pairwise comparison tests of mean points per shot will be performed for each cluster defined and for each season in the data to find significant differences in shooting efficiency.
The data used for this study is available from the NCSU Sports Statistics Club website in a CSV format.
