The use of statistical analysis has revolutionised sport in recent decades. Its success in baseball - as sabermetrics - was made famous by the book and movie (starring Brad Pitt) titled Moneyball. (In many ways, the book has been partly responsible for Kolkata Knightriders winning the IPL as its CEO, Venky Mysore, shaped his approach to picking the KKR squad based on principles outlined in it.) Now, the game of basketball could well go through a similar overhaul thanks to an unlikely source: a biomechanical engineering undergraduate at Stanford, prospective medical student, and Indian American from Houston named Muthu Alagappan. The magazine GQ has imaginatively named his concept 'Muthuball'.
Muthu began his deep dive into NBA positions one Friday last summer, during his internship with Ayasdi, by picking seven rudimentary statistics on Yahoo! Sports for every NBA player: points, rebounds, assists, steals, turnovers, fouls, and blocks. Then he adjusted them for playing time. Within three hours, he had a graphic with dense groups of color-coded nodes, representing last season's NBA players, connected by lines, expressing statistical affinities. The nodes were the players, the groups were the newfound positions, and the lines linked statistically similar players. This was his groundbreaking similarity network of NBA players. It looked like a postcard from a molecular biology convention.
"We expected to see five categories, or five flares, corresponding to the five positions of basketball. We actually found something much more interesting: We found that there's thirteen positions in the NBA," he said at the MIT conference. "And for each position, for the first time, I can topologically and mathematically define what it means to play that position. I can tell you which players in the NBA play which position. And I can tell you who epitomizes the position best."
There's additional (earlier) coverage over at Wired, NYT, and more. A video of his talk and associated slides are also available.
- Log in to post comments