March Madness brackets.
If you filled out an NCAA basketball tournament bracket last year, you probably didn't pick Duke to win it all. Few did. In fact, if you look back at all the ratings systems, analyses, polls and prognostications, only a handful of people thought Coach Mike Krzyzewski's team was the best in the nation. One of those guys was Ken Pomeroy.
Pomeroy, part-time college basketball number cruncher and full-time National Weather Service meteorologist, is a longtime stats geek who has developed a system to rate college basketball teams that is gaining prominence among mainstream analysts, players and even some coaches, especially after Duke's 2010 championship run.
"I try not to overplay the accuracy of it," says Pomeroy. "College basketball, like any sport — you can't predict it with absolute accuracy."
But he's getting close. Pomeroy has developed a ratings system based largely on offensive and defensive efficiency. Instead of looking at more traditional markers like a team's win-loss record, Pomeroy's ratings emphasize points scored per possession and points allowed per possession. Those numbers are then adjusted for the strength of the opponent, when the game is played (more recent games get more weight) and where it's played. (Home-court advantage is also weighted.) Pomeroy then filters those numbers through something called the Pythagorean expectation formula — don't make me explain it, but you can go here to see it — and voila! A set of numbers pops out that at times has been shockingly accurate in determining the best teams among the country's 345 Division I schools come tourney time.
Pomeroy's pre-tournament rating of eventual champ Duke as No. 1 last year wasn't his first. In his 2008 ratings, he also rated Kansas No. 1 (and with all four No. 1's making it to the Final Four that year, abiding strictly by his ratings, his bracket was practically flawless.) In 2009, Pomeroy had champ North Carolina at No. 2.
His analysis, of course, is not foolproof. But it's about as close as anyone has gotten, especially considering the unpredictability of the tournament. Unlike the NBA, teams have only one game to advance to the next round. And think of all the close calls, last-minute shots, unexpected injuries and unknown underdogs that make a deep run. If Butler's Gordon Heyward would have made his last-second, half-court heave in last year's championship game, ESPN analysts might not be talking about efficiency stats as much as they have been this year. (Disclaimer: Pomeroy had Butler rated 26th going into the tournament.)
Pomeroy's ratings originated years ago as he was watching Air Force, a team that played at an extremely slow pace. Most people credited their defense for their low-scoring games. But Pomeroy saw it differently.
"I always felt like their defense probably wasn't that good when you considered how many opportunities their opponent had," he said. "They milked the shot clock every single possession, and so they didn't have many chances to score points." Pomeroy crunched the numbers and found his hypothesis to be correct. It wasn't their defense that was keeping the score down. It was their offense.
After that, Pomeroy started his site, kenpom.com, and began rating teams in 2003. For a while, only college basketball junkies knew about it. And for the most part, it's still that way today. But it's growing in popularity. "This year there's kind of a tipping point with the TV guys," Pomeroy says.
One of those TV guys is Jay Bilas, one of ESPN's sharpest on-air commentators, who has been pushing hard for stats like Pomeroy's to get more attention. "I use them all the time," Bilas says. "It's one of the first things I look at, because they break the game down into greater detail."
Pomeroy's ratings aren't the only ones that fly below the radar. There are dozens of systems that rate teams using multiple metrics, including Joel Sokol's LRMC Bayesian results (using only basic scoreboard data, like which two teams played, where they played and the margin of victory) and Sonny Moore's Power Ratings (using wins, losses and scoring margin). And then there's Nate Silver's fivethirtyeight blog at the New York Times' site. Silver is normally concerned with political calculations, but come tournament time, he painstakingly fills out his own bracket, aggregating all the ratings systems above as well as seeds and poll numbers.
The father of many of today's more modern ratings systems is often considered to be USA Today's Jeff Sagarin, one of the first statistical analysts to find a proper way of combining scoring margin and strength of schedule into one overall number that can be used to rate teams. Sagarin's reliable numbers are often similar to Pomeroy's, although he says he doesn't go to his site often because "it just clutters up my eyes looking at [his numbers]."
He's got a point. Even though Pomeroy claims that his analysis is a simple process, it can take a while to determine just how he churns out his numbers. (He even has a column for "luck" that isn't actually computed into his ratings but is an attempt to quantitatively analyze teams based on the breaks they get in games.)
Pomeroy's results, however, speak for themselves.
During ESPN broadcasts, talk of offensive efficiency has slowly crept into how analysts describe teams, Bilas being the most prominent public cheerleader. So far, however, the NCAA selection committee doesn't appear to be using Pomeroy, Sagarin or any other more recent ratings systems to seed teams. Instead, it still relies on metrics like the old-school RPI (Ratings Percentage Index), which strongly emphasizes strength of schedule.
RPI is anathema to guys like Bilas, who questions why the committee uses a "blunt instrument rather than a precision tool" to pick seeding. But over the years, as newer ways to rate teams gain prominence, that may change.
So now that the tournament field is set, who does Pomeroy like to win it all? Currently, powerhouse Ohio State has the top billing in his ratings.
"I don't want to sell the notion too hard that whoever is No. 1 in my rating is going to win the national title," Pomeroy says. "But I'm feeling pretty good. It doesn't seem people are as excited about them as Kansas right now. But if I was going to pick a team to go all the way, I'd pick Ohio State."
But even Pomeroy doesn't want people blindly abiding by his ratings. In fact, he doesn't even do that.
"I don't go with my stats in picking a bracket, because that's boring," he says. "The tournament's about upsets. And sometimes I go with my gut, but my gut's horrible. Most of the time I'm wrong, just like everybody else."