For Pro Teams, This Information Really Computes


It was June 11, 1996. The Seattle SuperSonics were down, three games to none, to the Chicago Bulls in the NBA finals. It seemed that only a miracle could save them from elimination.

At the same time, Inderpal Bhandari of IBM was preparing his keynote speech at a conference on data mining in Santa Clara, Calif.

The India-born computer engineer had been analyzing the championship series. And his data had led him to a surprising conclusion.

“Frank Brickowski,” Bhandari said, with absolute conviction, to Steve Hellmuth, NBA vice president for operations. “Frank Brickowski is the key.”


Hellmuth was perplexed. Brickowski was the SuperSonics’ backup center, a player who had not scored more than a few points nor grabbed any significant rebounds in any of the first three games. How could he possibly be the key to turning the series around?

“He doesn’t show up in the box score,” Bhandari explained, pointing to the figures on his computer monitor. “But when he’s on the court, Seattle outscores the Bulls. It’s all right here in my data.”

After sharing his observation with his conference audience, Bhandari faxed his observation to the Seattle coaching staff. Brickowski was put into the starting lineup, and helped the SuperSonics win the next two games before Chicago took the sixth to win the series and the NBA title.

Advanced Scout is the name of a data-mining program that Bhandari developed as a tool to assist NBA coaches. It was through this program that Bhandari was able to recognize the pivotal role Brickowski played for the SuperSonics against the Bulls.


“Advanced Scout allows our coaches to quantify the intangible,” says Hellmuth, who oversees the NBA’s collection of 2.5 million play videos.

“It allows you to ask questions that you don’t know how to phrase. Who are my best five players on the court in the third quarter against the Houston Rockets? Who are the best shooters in the league with less than a minute to play and the score tied? This is how you truly find your buzzer beaters.”

Bhandari became a basketball fan by chance.

“I was looking for a way to make data mining available to the non-technical user,” says Bhandari.

“I was also looking for an application that would create the broadest possible impact. One day I read about how Pat Riley was interested in statistics, and that he had asked one of his assistants to create a database for him when he was coaching the New York Knicks. I figured that he might be interested in the type of analysis he could obtain through data mining.”

Riley was more than interested. His team had just lost to Houston in the NBA finals, and he was eager to understand why. He assigned Bob Salmi, one of his assistant coaches, to work with Bhandari. Tom Sterner, an assistant coach with the Orlando Magic, also jumped on board.

Hellmuth, who oversees the NBA’s entertainment division, was quick to see the potential in adapting data mining to the specific context of the NBA.

“Data mining helps you to identify situations that you might not normally see,” says Sterner, a veteran coach who holds a master’s degree in sports administration and computers from Temple University in Philadelphia. “A good coach may have a feeling what his best rebounding, shooting, or defensive team is. The technology is there to confirm his intuition.”


By nature, basketball may be no better suited for data mining application than such sports as baseball, football, or soccer. But the NBA had already laid a solid technical foundation. Since becoming an official NBA sponsor in 1991, IBM has provided NBA teams with computers, and assigns three people with laptops to keep statistics at courtside at every game. Every shot, pass, block, foul, and substitution is logged and recorded.

Yet a set of data, no matter how comprehensive, does not necessarily produce relevant information. Even a few years ago, coaches had to view hours and hours of tape in order to glean a valuable insight, or to find a piece of film that illustrated a particular pattern or situation.

Today, when Advanced Scout points out a problem, it also provides the exact times that problem occurred. The relevant digital video clips can be viewed in an instant.

“I think it was [Indiana University Coach] Bobby Knight who said that statistics accuse and video indicts,” says Hellmuth, who intends to use Advanced Scout to help edit overflowing play video archives from which the NBA makes its highlight and promotional films.

As of last September, when the NBA held its annual technology conference at Secaucus, N.J., 15 NBA teams had agreed to begin implementation of Advanced Scout in their coaching systems. Since then, three more teams have expressed interest in the application for this season.

Because of the success of its program in basketball, IBM has recently joined with the NHL to adapt Advanced Scout to its needs.

“The general questions regarding the application are going to change,” says Bhandari. “It probably won’t make much sense mining on goals, because there are so few. And it doesn’t make much sense to search for the ideal combination of players, because the players come on and off the ice so frequently.

“We will probably have to center our analysis on quality shots, puck control. We will work with a few coaches and assistant coaches to understand something about the game, and about how decisions are made.”


It remains to be seen whether state-of-the-art intelligence and technology can transform an also-ran into a winner. Last season, Sterner’s Orlando Magic met the Chicago Bulls in the NBA’s Eastern Conference finals. The Magic lost Games 1 and 2. But Advanced Scout showed Sterner that the Magic tended to do its best against Chicago with Penny Hardaway and Brian Shaw in the backcourt.

Sterner relayed the information to Coach Brian Hill, and Hill started Shaw and Hardaway as guards in Game 3. The Magic played Chicago even in the first half. But in the third quarter, Chicago’s gifted Michael Jordan simply took over the game, and the Bulls went on to sweep the series.

“There is no compensating for talent,” Sterner says. “If you have superior talent, you win. But in a situation where one team’s talent is equivalent to another’s, this technology can provide you with the difference between winning and losing.”