|Data Mining in the National Basketball Association
When the Orlando Magic were devastated in the first two games of the 1997 National Basketball
Association (NBA) Finals against the second-seed Miami Heat, the team's fans began to hang their heads in shame. But fortunately, the Magic had another trick up their sleeve. A data mining application developed uncovered a secret buried beneath the layers of statistics collected at every game.
The application, Advanced Scout, is specifically tailored for NBA coaches and statisticians. Advanced Scout showed the Orlando Magic coaches something that none of them had previously recognized. When Brian Shaw and Darrell Armstrong were in the game, something was sparked within their teammate Penny Hardaway, the Magic's leading scorer at that time. Armstrong received more
play-time and hence, Hardaway was far more effective. The Magic went on to win the next two games
and nearly caused the upset of the year. Fans everywhere rallied around the team and naysayers quickly replaced their doubts with season ticket purchases for the following year.
Coaches, like business executives, carefully study data to enhance their natural intuition when
making strategic decisions. But unlike business, the direct results of coaching decisions are played out under the eyes of millions of fans, and wrong calls can turn a team's fans against it, leading to lower ticket sales and possibly a vacancy in the head coaching position. By helping coaches make better decisions, data mining applications are playing a huge role in establishing incredible fan support and loyalty. That means millions of dollars in gate traffic, television sales and licensing.
Before these data mining applications, some teams, such as the Orlando Magic, began developing business intelligence software to find patterns in the piles of game data that the coaching staff collected during play. But with an average of 200 possessions a game and about 1,200 games a year, the sheer volume of statistics was overwhelming, and the applications produced only basic results--the kind of statistics anyone could find in a local newspaper.
During the course of each game, members of the NBA's Game Stats program manually enter game statistics into laptops. This data is then uploaded to a server. Coaches can log on to the data mining application before, during or after a game to download this public data and merge it with the private data that each team collects independently on its laptops or PCs.
Using the data mining software, coaches can drill down into a vast array of statistics and data
and unearth comprehensible patterns that were previously hidden among seemingly unrelated statistics.
Coaches can ask which players are most effective in correlation with time and the opposing players. Coaches are able to get, in real time, statistical evaluations that allow them to put in the very best players for specific points in the game. The application really helps coaches understand the relationships
among the combinations of players on the court, changes the way they coach their teams, and helps them make more effective decisions."
Data collected by the data mining application is also stamped with a universal time code. This means that when coaches see an intriguing pattern, they can get the exact moment in the game where it occurred and instantly cue this up on videotape. Coaches used to spend weeks scouring the tapes, but now they can instantly pull up a key spot in the game that, before using data mining, they probably would not even have known to look for.
While coaches currently have a robust tool with them at courtside to optimize player line-up,
data mining functionality will soon be expanded to include analyzing the effectiveness of
specific plays that teams have designed. Coaches are going to be able, right from courtside, and in real time, to ask the application which play will be the most effective relative to the time elapsed and the specific combinations of players on the court.