USC Researchers Measure World Series Player Buzz

IBM and USC researchers are measuring opinion using "positive" and "negative" tweets.


As the World Series kicks off, fans across the nation will be watching the Texas Rangers' and St. Louis Cardinals' loaded lineups pile up runs. Meanwhile, researchers at the University of Southern California will have their eyes on a different statistic: tweets per minute. 

Using new IBM social media tools, the group is trying to determine—in real time—fan favorite players by analyzing the number of tweets with "sentiment," either positive or negative. The software calls on a dictionary of 1,800 sentiment-charged words and analyzes whether tweets are positive or negative. 

If the World Series games were determined by Twitter buzz, the Rangers would be smashing the Cardinals—the night before the deciding games of the Championship Series, the Rangers garnered some 56,600 sentimental tweets, five times the Cardinals' 11,500. 

With 1,573 tweets before the final game of the NLCS, longtime Cardinals pitcher and fan favorite Chris Carpenter, who shut down the Philadelphia Phillies in the National League Division Series' deciding game, got the most buzz of any player. Only about 60 percent of the tweets were positive. 

USC professor Jonathan Taplin says that his students were able to estimate TV ratings for the NLCS games based on positive sentiment—potentially important data for team owners. Owners could also determine which players have the most marketability. 

"We understood that the people who were leading in terms of fan sentiment and everything else were not highest paid players in the world," Taplin says. "You don't have to have the most highly paid player to be the most popular." 

Rod Smith, vice president of emerging internet technologies at IBM, says the company is working with journalism schools and companies to make real-time analyses of consumer sentiment, something USC has already started working on. 

Taplin says his classes have analyzed movie opening success with a program they call Film Forecaster. 

"Studios said that it didn't help them knowing that "Cowboys and Aliens" was going to tank two weeks before it opened," he says. But Taplin, a former film producer, says film and television studio executives have shown interest in using the technology to measure ad campaign success. 

Real-time analytics could even one day take over conventional political polling—during the Republican debate Tuesday, the program was crunching more than 5,000 tweets per minute. 

"You've got all these opportunities to make this information available to a broad range of constituents," Smith says. 

For now, the researchers are focused on baseball—they'll be trying to see if positive Twitter buzz aligns with player performance. 

"We're going to track every player in terms of positive and negative sentiment and see if the fans agree with the sportswriters in terms of who was the most valuable player," he says.