Moments after his team routed Connecticut in the American Athletic Conference championship game Saturday, Louisville coach Rick Pitino explained, with one of the biggest clichés in a month full of them, why he felt his team deserved a No. 1 seed in the NCAA Tournament.
"I think we pass the eye test," he said as the confetti fell onto the FedEx Forum floor in Memphis.
The problem for Pitino, whose red-hot Cardinals were assigned a No. 4 seed in the Midwest Region, is that the eye test isn't centered only on the court these days. It's an examination that focuses on figures such as adjusted defensive efficiency (AdjD), effective field-goal percentage (eFG%) and points per possession (O-PPP), a new wave of advanced basketball metrics used to identify cracks in a team before they are exposed on the tournament stage.
Welcome to a new kind of March Madness, where the names Ken Pomeroy and Nate Silver — statistical gurus who have found new ways to harness data — are as recognizable to the sport's insiders as Mike Krzyzewski and Jim Boeheim.
With the NCAA Tournament tipping off in earnest Thursday, what does it all mean to you, trying to win your office pool? Simply put, there are more numbers than ever being crunched in an attempt to forecast the most unpredictable sporting event in this country.
That hasn't stop millions of Americans from taking a stab at picking a perfect bracket, which this year could net you a cool $1 billion of Warren Buffett's money through the Quicken Loans Bracket Challenge. That total would mark just around one-third of the estimated $3 billion waged in office pools, based on a Bloomberg News study.
But before making a mad dash for cash based entirely on new wave statistics, beware. A freshman trying to make a clutch free throw in front of thousands of screaming fans can still wreck your bracket.
"I have to remind people that players actually generate the data," said statistician Jeff Tiernan, who runs the college basketball website bracketscience.com and has studied the NCAA Tournament for 22 years. "The data doesn't move the players around. They still have to go out there and play the games. These are 18- and 19-year-old kids. And therein lies the chaos."
Alok Pattani spends his days using numbers to explain that chaos. An analytics specialist at ESPN, Pattani helped create the network's Basketball Power Index (BPI) two years ago. It's a schedule-adjusted rating system that takes into account factors such as margin of victory, pace of play and the absence of key players to gauge a team's overall strength.
In its first two years of existence, the team with the highest BPI ranking entering the NCAA Tournament — Kentucky (2012), Louisville (2013) — won the championship. This season, that team is Arizona.
"The margin in (their) losses was close, they were all (away from home) against good teams. You take all of that into account and judge it against what an average team would do in that situation," Pattani said.
Pattani is quick to point out that BPI, co-created by renowned statistical expert Dean Oliver, is just one of many new rating systems. Following any one system won't guarantee results anymore than choosing a stock does. As with any statistical analysis, a larger sample size produces more certainty.
Still, numbers can help you. The 12-over-5 seed in the second round of the tournament has become more of a trend than a surprise. Since the tournament field was expanded to 64 teams in 1985, a No. 12 has defeated a No. 5 team 40 times, including three times last year.
"So you say, 'OK, which one should I pick," Pattani said. "I should probably pick a strong 12 against a weak 5. So I don't want to pick one where the 12 is more like a 14 and the 5 is more like a 3. So you can pick upsets that can be informative in some sense."
If you use BPI, as an example, Harvard (No. 38) is the highest-rated 12-seed, ranked seven spots ahead of Texas, a No. 7 seed.
So, is making the right upset picks the best way to win your pool? Is there a single advanced metric that can predict a team's success? In short, Pattani said, the answer is no. To paint the most accurate picture of a team, no one category tells the entire story.
For example, Wichita State, the No. 1 seed in the Midwest Region, has taken criticism for a strength of schedule that ranks well below that of many top tournament teams. But disregard the 34-0 Shockers at your own peril. They rate well with statistical experts such as Pomeroy because of high marks in adjusted defensive and offensive efficiency, metrics used to determine how many points a team would score or allow per 100 possessions against an average Division I opponent.
"It's not one of those things where it's like, 'here's strength of schedule; here's point margin; here's win-loss,' " Pattani said.
Past predicts future
Learning statistical history can be helpful, too. Tiernan uses past tournament results to formulate some of his predictive bracket models. In one example, he created a list of eight criteria — factors such as averaging more than 73 points per game and having an average scoring margin of at least seven points per game — that all of the past 13 NCAA champions possessed. Using those parameters, only six teams — Arizona, Wisconsin, Duke, Michigan, Kansas and Villanova — have a shot at this year's championship. Of that group, only Arizona is a No. 1 seed. So, while most analysts are predicting a wide-open tournament field with a dozen or more teams capable of winning the title, his research doesn't necessarily back that up.
Tiernan said he'll run as many as 10 or 12 different models based on different criteria in order to gauge a wide range of champions in different scenarios.
"Because to me, that is reality," Tiernan said.
And, as much as it may hurt the analytical side of him to admit it, Tiernan's "from the gut" model of predicting often does as well as any of statistical-based analysis. As soon as the bracket comes out, he fills it in before examining the numbers.
"I think it's the Malcolm Gladwell maxim that he forwarded in "Blink," Tiernan said. "That sometimes in the matter of complex decision-making, our instincts are better than our overthinking."
Still, Tiernan encounters plenty of fans who rely on his numbers. And, with $1 billion to be made, more fans than ever are looking for an edge.
Predicting the field
A look at what some top data crunchers are saying about the NCAA Tournament:
Nate Silver, FiveThirtyEight.com: Forecasts No. 4 seed Louisville as having the best chance of winning the tournament, at 15 percent, followed by Florida (14%) and Arizona (13%). He gives Colorado a 28% chance of winning its first game, against Pittsburgh, and less than a 1 percent chance of reaching the Elite Eight.
Ken Pomeroy, of KenPom.com: Gives Arizona the best chance of winning the title at 16%, followed by Florida at 13%. Pomeroy's computers like Arizona because it is ranked first by a wide margin in adjusted defensive efficiency, a metric that measures how many points per 100 possessions a team would surrender against an average Division I team. (He gives CU a 1-in-5,169 chance of winning it all.)
ESPN's BPI-based projection: The network's ratings system, created by Alok Pattani and Dean Oliver, gives Arizona a 21% chance to win the title.
Nick Kosmider, The Denver Post