<
>

Giant Killers: This 'random' NCAA field isn't so hard to predict

Matt Cashore/USA TODAY Sports

This story appears in ESPN The Magazine's March 28 MLB Preview Issue. Subscribe today!

Stop us if you've heard this before: This has been one crazy season in college basketball. By now you know that six teams have topped the AP poll entering the NCAA tournament. You've heard that teams ranked in the top five have lost to a record 21 unranked opponents. And it's not just you who feels a bit lost when it comes to filling out your brackets -- even analysts and media types are as confused as presidential super PAC donors.

With no established favorites -- all together now: "Parity!" -- this year's field seems wide open and virtually unpredictable. But ... not exactly. It turns out that some teams with chaotic results in the regular season aren't actually that hard to predict: They're going to be, well, chaotic (in the NCAA tournament).

That's what we learned when, as part of this year's 10th annual Giant Killers project, we asked: After a season that seems so random, which teams have the potential for the craziest results -- particularly in a one-and-done format, with its own inherent randomness? In short, we went in search of the predictably unpredictable. You see, an unpredictable high seed can't comfortably be expected to lock down even an inferior opponent, which makes that team a vulnerable Giant. On the other hand, low seeds with a genuine capacity to surprise make terrific Killers.

To measure chaos, we used a process developed by Tim Chartier, associate professor of mathematics at Davidson College and chief researcher at Tresata Inc., a data intelligence company in Charlotte, North Carolina, with help from Kevin Hutson, associate professor of mathematics at Furman University. We started by ranking all the teams in each tournament since 2010 from 1 to 68 (1-65 in 2010). We used the Massey method, which employs linear algebra to calculate the strength of all the teams in a set, adjusted for the strength of their opponents. And we excluded close games to focus on results we could be sure were nonrandom. Then, for each year's field, we looked at how the rankings changed when we removed the results of any one team.

If a team makes sense -- if it beats lesser opponents and loses to better ones -- the overall rankings won't change much when it is removed. It is a predictable team. But if a team's results carry a lot of surprise information, there will be considerable change between the before and after rankings. It is an unpredictable team.

Consider Georgetown in 2009-10. The Hoyas were a fine team, but even before the tournament, they seemed a bit all over the place: They won eight straight to open the season, then lost four of their final six regular-season games (they finished 23-10), then made it to the Big East final. They beat Duke and Syracuse, both ranked in the top 10, but lost to unranked Rutgers and South Florida (teams they beat by a combined 45 points in their other two matchups). Sure enough, according to our new analysis, they were the most chaotic team in the country that season. And guess what: Georgetown carried that unpredictability into the NCAA tournament, where the Hoyas were upended in the first round by Ohio, a 14-seed seeking its first tourney win in 27 years.

There's so much more where the Hoyas came from. Saint Mary's, the second-most chaotic team in 2009-10, won two tournament games as a 10-seed, including a Giant Killing -- an upset with at least a difference of five seeds between the teams -- of 2-seed Villanova. Wichita State, No. 1 in randomness in 2011-12, went down to VCU in a 5 vs. 12 upset. San Diego State, 2012-13's most random, fell to 15-seed Florida Gulf Coast. Last year the tournament's 10 most chaotic teams included UCLA, Virginia, Villanova and Baylor -- one Killer and three slain Giants. Overall, the top 15 teams in each year of our chaos ratings have accounted for nearly half (29 of 59) of the Giant Killings since 2009-10.

So what does that mean for this year's tourney? In a season filled with chaos, a host of teams deserve close scrutiny. Start with three flagship programs: Michigan State, North Carolina and Kentucky. The Spartans and Tar Heels are among the favorites to win the title, and deservedly so. But Sparty's midseason swoon of three straight losses to Iowa, Wisconsin and Nebraska triggered alarms in our model. UNC -- with wins against Maryland, Duke and Miami but also losses to Northern Iowa, Texas and Notre Dame -- has been tough to peg all season. Indeed, it shouldn't surprise anyone if a low seed packs the lane against Carolina's 31 percent 3-point shooters and rides a barrage of bricks to a major upset. Kentucky falls into that bucket as well: The Wildcats are rated as the eighth-best team, according to KenPom.com, but have losses against Auburn (211th), Tennessee (110th), LSU (82nd) and UCLA (72nd).

At the other end of the spectrum, you'll find a familiar Killer: VCU. Two years ago, then- assistant Will Wade left the Rams to take the head-coaching job at Chattanooga and brought Shaka Smart's "Havoc" system with him. He rebranded it as, wait for it, "Chaos." Now Wade is back at VCU and has the Rams playing their typical high-risk, high-reward style. They lost nine regular-season games, but their pressure D (22.7 defensive turnover percentage, eighth in the NCAA) can topple any shaky Giant. And if you're looking for an even deeper sleeper, try Stony Brook. The Seawolves pushed Vanderbilt -- KenPom's 23rd-ranked team -- to the limit in an overtime loss and then fell to two teams outside of the top 150 (Western Kentucky and Vermont). With 6-foot-8, 260-pound Jameel Warney hitting the offensive glass (4.4 orpg), SBU excels at creating extra possessions, the hallmark of a strong Killer.

So fear not the randomness that has engulfed college basketball this season. Embrace it. And may the upsets fall in your favor.