Team A: Video data extraction
Alexander Bendeck (Computer Science, Statistics) and Niyaz Nurbhasha (Economics) spent ten weeks building tools to extract player and ball movement in basketball games. Using freely available broadcast-angle video footage which required much cleaning and pre-processing, the team used OpenPose software and employed neural network methodologies. Their pipeline fed into the predictive models of Team C.
Team B: Modeling basketball data: offense
Anshul Shah (Computer Science, Statistics), Jack Lichtenstein (Statistics), and Will Schmidt (Mechanical Engineering) spent ten weeks building tools to analyze offensive play in basketball. Using 2014-5 Duke Men’s Basketball player-tracking data provided by SportVU, the team constructed statistical models that explored the relationship between different metrics of offensive productivity, and also used computational geometry methods to analyze the off-ball “gravity” of an offensive player.
Team C: Modeling basketball data: defense
Lukengu Tshiteya (Statistics), Wenge Xie (ECE), and Joe Zuo (Computer Science, Statistics) spent ten weeks building tools to predict player movement in basketball games. Using SportVU data, including some pre-processed by Team A, the team built predictive RNN models that distinguish between 6 typical movement types, and created interactive visualizations of their findings in R Shiny.
Team D: Visualizing basketball data
Shixing Cao (ECE) and Jackson Hubbard (Computer Science, Statistics) spent ten weeks building visualizations to help analyze basketball games. Using player tracking data from Duke basketball games, the team created visualizations of gameflow, networks of points and assists, and integrated all of their tools into an R Shiny app.
This project aims to analyze assessment and performance data collected from baseball players to make predictions about baseball performance based on vision and physical abilities. We use hierarchical regression analyses to identify characteristics that correlate with batting performance in order to inform scouts about the likely production of developmental prospects. The final product is an application that uses an athlete's assessment results to produce performance summary graphs for the individual compared to other athletes and inferential models for the relationships between assessments and performance.
Watch the team's final presentation (on Zoom) here: