WHAT MAKES A BASKETBALL game “explosive?” Commentators know it when they see it. But for Anil Timbil ’20, a student developer at Google Cloud, the answer is in the data.
This season, Google Cloud partnered with the NCAA to provide real-time March Madness analysis, including game predictions that were featured in TV coverage of the tournament. Computer science major Timbil was one of around 30 students charged with creating these insights by crunching a decade’s worth of play-by-play data that the NCAA shared with the internet giant.
“This was the perfect opportunity to combine my interest in basketball and intensive computer science and statistics studies,” said Timbil, a fan of the sport from an early age.
The “explosiveness” question first came up last January, at the Google Cloud & NCAA Hackathon at MIT. There, Timbil and team wondered whether it was possible to quantify the commentators’ intangible claims of a team’s explosiveness. They turned to the data to find out.
First, they focused on “whether dunks had any demonstrable effect on the energy or momentum of a team—something often assumed to be true, but rarely (if ever) exposed with data.”
The short answer: they did.
“Dunks actually contributed to about a 15 percent increase in game acceleration—a much bigger effect than we’d anticipated,” Timbil recalled.
But as good as that insight was, it only indicated so much about a team’s performance. More interesting was what this insight said about explosiveness over the course of a game.
“We wanted to tie explosiveness to some kind of measurable increase in score differential over a specific period of time,” Timbil explained—in other words, how quickly one team outscores another during a run. The students defined a run as one team scoring at least 12 points and the other at most five.
“All together, we considered explosiveness to be the product of weighted speed of scoring and opponents’ stopping power, plus a weighted value for shot accuracy,” Timbil said. “Lots of points in a run is better than a short run with a slightly higher rate of points over time, but we still wanted to be sure to value faster runs as more explosive.”
The last step was using these new data methodologies to rank the explosiveness of this year’s teams. These rankings became game predictions featured in Google commercials screened during the tournament. One even featured Timbil.
The key to success in the data and statistics fields is persistence and diligence, according to one of Timbil’s teachers, Associate Professor of Computer Science Christine Chung.
“The way Google is so publicly highlighting Anil’s team’s work during a television broadcast that reaches millions of homes indicates that it must be world-class,” Chung said. “These novel techniques for analyzing huge amounts of data to pin down a nebulous predictive factor like ‘explosiveness’ can and should be applied to other sports in the future, as well.”
This April, Timbil was also one of 35,000 attendees at the Google Cloud Next conference in San Francisco. Timbil and the other student developers received a shout-out from Google CEO Sundar Pichai during his keynote address with Timbil appearing on-screen behind Pichai sporting a Conn hoodie.
“It was really empowering to represent the college among the most competitive universities in the United States,” said Timbil, who will be continuing his data science work at a New York City internship this summer.