Earlier this week, we read with interest in The New York Times that Davidson College’s math department is helping its men’s basketball team win games. Since this combines three of our favorite subjects — sports, math and using math to win at sports — I asked Tim Chartier, a professor in Davidson’s mathematics and computer science department, to elaborate on the unusual partnership between his department and the Wildcats.
He was understandably reluctant to spill too many details — particularly regarding how the math department analyzed the data about Iowa, Davidson’s round-of-64 opponent in the NCAA tournament Friday, or Gonzaga and North Dakota State, one of which Davidson will face if it reaches the round of 32 and both of which he’d already started scouting. But Chartier did shed a little more light on the math behind the surprise Atlantic 10 champs. And it sounds a lot like what we do at FiveThirtyEight, except with a goal of winning basketball games instead of writing articles about winning basketball games. Here are edited excerpts of his reply:
So, what do we offer the coaches?
Heat maps: Two members of Cats Stats, as we call our group, track the games for heat maps. They mark who takes a shot and estimate the location by clicking a computer program that I wrote. Then, as you see, regions of the court are marked as hot and cold regions.
Two members of the group met with coaches in January and went through heat maps not just at the team level but player level and then various combinations of players. The coaches found this very helpful and discovered things that led them to analyze player combinations with similar tendencies that they then recognized.
Lineup efficiencies: Here we show Dean Oliver’s Four Factors for offense and defense for every five-man lineup and the total amount of time that lineup has been used. Then, we also found various subsets of lineups, as well, like guard trios and forward pairs. For example, we might see that two big men are most efficient/inefficient when in the game together.
Personal scouting: Here we produce a detailed breakdown of players’ tendencies and how well they performed in each aspect of the game. Here no new statistics were created. We presented data and numbers so that the coaching staff could easily implement the reports into their game plan. It could range from things like whether or not a player liked to shoot off the dribble and how effective they were in those situations to things like how effective a player might score driving left out of isolation situations. Said another way, we look for the tendencies in the numbers, especially outlier numbers. We think of these as creating data points or dots. Then, we dig into the video that accompany these stats on Synergy and figure out the story/tendency behind the number. That gives context and, in a sense, connects the dots. For example, if a player drives to the right 80 percent of the time and scores much better than going left, that is important. But, then the video can help see that his first step when driving is quicker going right than left, allowing him to beat his man more often.