First year Pitt Statistics PhD student Marc Richards recently won a Data Competition and will be presenting at the RIT Sports Analytics Conference on Saturday, September 14th. The RITSAC Data Competition was sponsored by ICEBERG Sports Analytics. Using hockey data such as shots and passes, and locations of each player on the ice throughout the game (tracked at a rate of 10x/second), ICEBERG challenged participants to develop new ways of analyzing hockey performance, strategy, and tactics through player tracking data. 

Schedule of events and Data Competition Description can be found here:

https://www.rit.edu/science/sports-analytics-conference

Marc's winning submission was an Expected Completed Pass Model. With a detailed introduction/abstract below:

Expected Goals (xG) have become a stable in the hockey analytics community over the past few years. While Goals are the ultimate end goal, they often do not occur without a pass. To better understand players ability and the difficulty in certain pass attempts, I explored the concept of an Expected Completed Pass Model (xP). Much in the same way of an Expected Goals model, the purpose of the model is to estimate the probability of success. Except, in this case, a success is defined as a pass that is completed rather than a shot that goes in the back of the net. For example, if Player A passes to Player B, we can assign a likelihood of the pass being completed. Through the use of Player and Puck Tracking data provided by Iceberg Hockey analytics, one can identify when passes occurred, who they were targeted for and whether or not they were completed. Given the binary nature of our response variable, whether or not a pass was completed, we utilized a gradient-boosted ensemble of decision trees model.