The NCAA basketball tournament brackets are set: Florida, Arizona, Wichita State, and Virginia were awarded the four regional top seeds. Let the office pools commence! For help, we sought advice from Ken Massey, one of the foremost minds in quantitative sports analysis. A math professor at Carson-Newman College in Tennessee, his Massey Ratings have been around since the 1990s—the Bowl Championship Series has incorporated his college football ratings since 1999. His rating system focuses on who beat whom, when the game was played, and where. It has a strength-of-schedule element that Massey claims is “better than RPI.” Using the same math he used in football, Massey has expanded his approach from football to other sports, including college hoops.
Massey’s algorithm doesn’t use detailed sport-specific box-score analysis that somebody like Ken Pomeroy might. All those details roll up into the final result, he says. The effect of who wins is all that matters, and his rating system follows that logic. A team that wins two games by 3 points each is rated higher than a team that wins one game by 20 and loses the next by 1.
Margin of victory is secondary, he says. It also has diminishing returns. The difference between a 30-point win and a 20-point win has less value than the difference between a 12-point win and a 2-point win. “Wichita State has dominated this year,” he says, because of its huge margins of victory while going undefeated, despite playing an easy schedule.
Here’s the bad news. Massey’s work suggests that no matter how much extra data people add to their models, the probabilistic forecasts reach a ceiling in their ability to accurately predict games. Historically, favorites win about 70 percent of the time, and that’s about as high as anybody is going to reliably get. Extra details about the game’s statistics are only marginally helpful in predicting outcomes. Massey believes these details would have more importance to coaches and teams, who could use the additional data to help their in-game strategy—but it wouldn’t do much for regular people trying to win a bracket pool.
Specifically for 2014, his ratings (through games as of March 15) made Arizona the favorite, with a 14.5 percent chance to win the title. That’s not very good: The top team typically has a 22 percent to 25 percent chance to win it all. Some years, favorites have had chances higher than 30 percent. Having a top team in the mid-teens will only make it harder to fill out a winning bracket: Even the favorite is kind of a long shot.
Massey believes his ratings alone can help you do pretty well in a typical pool. In order to win the pool, you’re going to have to take chances. Two teams stand out in particular: Oregon, which is seeded seventh in the West Region, started strong this year and then faltered. If the Ducks can “right their ship,” Massey says, they’re poised to make a deep run. He also likes Virginia Commonwealth, a No. 5 seed in the South Region. The Rams reached the Final Four three years ago, and his data suggest they have a shot again this year.