Mix sports analytics, sunny weather, a beautiful campus, interested students, and great speakers and what do you get? The 2016 Carolina Sports Analytics Meeting (CSAM) of course! CANA Advisors, along with ESPN, Furman, and National Amateur Sports, sponsored this event. The plenary speakers were David Kaplan, Director of Analytics for the Charlotte Hornets, and Peter Keating, writer for ESPN the Magazine. Other speakers and attendees ranged from high school and college students to directors of analytics for professional teams. CANA presented a NHL sports analytics project and how technology was overwhelming sports with a tidal wave of data. A lot to pack into a one day conference.
David Kaplan’s talk kicked off the conference. His presentation centered on how to make analytics actionable. He told a story that occurred during his time as an intern with the Portland Trailblazers. He told an assistant coach that his analysis indicated that a certain player was shooting poorly in the third quarter. The coach turned to David and told him "Well what do you want me to do, tell the player to shoot better?". Analysis must support a decision to change behavior at either the management, coaching or player level. In the NBA, analytics is applied to defense by predicting what side of a defender the ball handler prefers and is more successful at scoring. With this knowledge the defender can attempt to force the ball handler to cut to the non-preferred side. David also stressed the fact that basketball is not a sport that major adjustments can be made by either coaches or players during the game. The pace of play and the short break lengths (timeouts, breaks between quarters, etc.) force small changes. This means the analytics products are delivered before the game and are rarely updated during the game. Another interesting point was that David was surprised how much writing is required by his position. He communicates almost all of his analysis by writing reports and providing them to coaches and players. This consumes a large portion of his day.
A popular event during the lunch break was the student poster presentations. This event gave students an opportunity to present and discuss their research. Subjects ranged from the effectiveness of randomizing softball pitches to ranking NBA players to a new method to find the overall best lifter at a power-lifting meet. Everyone learned something new, and the students got valuable feedback on their research.
Peter Keating closed the conference with an interesting talk on predicting NCAA Tournament upsets. He has worked the past few years with professors from both Furman and Davidson Universities on a model to predict when these upsets will occur. The model uses logistic regression techniques, and he reviewed the advantages and disadvantages of using this method. He recommended exploring other methods to compliment predicting these unlikely events such as common opponent neighborhood network analysis. Since all teams in the NCAA don't play all other teams, this method would leverage graph theory to tease out the probabilities using a common opponent approach (Team A plays Team B and Team B plays Team C, therefore there is a connection between Team A and Team C). He also stressed the importance of mathematical methods to find similar games. This would include parameters such as seeding of the two teams and the classification of what type of giant (favorite) and giant-killer (underdog). The presentation highlighted some important points in identifying rare events that could be applied to other areas outside of sports such as risk analytics.
Overall the conference was very successful. The smaller size of the conference ensures that everyone gets to interact with speakers and with each other. Also, April in Greenville, SC is beautiful. There were no less than two weddings, one 5K run and three other events occurring at the same time on the Furman campus. Already looking forward to next year!