Giant Killers: How it all works
An explanation of the methodology behind our upset-picking model
Welcome to the fifth edition of Giant Killers, our annual demystification of the craziest, most unpredictable event in sports: the NCAA tournament. Our goal is to uncover the huge upsets that you will remember forever and will wreck your cousin's brackets. (Think Princeton over UCLA, Santa Clara over Arizona, Arkansas-Little Rock over Notre Dame.) And here's the method to our March Madness.
First, some definitions. A Giant Killer is a team that beats a tournament opponent seeded at least five spots higher in any round. Squads from the six BCS conferences as well as Butler, BYU, Gonzaga, Memphis and Xavier (because of current or historical success) are ineligible -- those teams aren't sneaking up on anybody. In addition, any team that earns a top-five seed (which likely will include Northern Iowa, New Mexico and Temple this year) mathematically can't be a Giant Killer and is thus classified as a Giant.
A Slain Giant, on the other hand, is simply a team that loses to a Giant Killer. When No. 7 seed UNLV beat No. 10 seed Georgia Tech in 2007, that (obviously) was not a GK victory. But when the Runnin' Rebels went on to beat Wisconsin, a No. 2 seed, in the second round that year, that was a Giant Killing. The same goes for a No. 9 seed that upsets a top seed in the second round, and so forth. Since 2004, 24 Killers have beaten 28 Giants (a few, such as George Mason in 2006, slayed multiple victims).
Every year, we comb through dozens of statistics for every program in the country to find underdogs poised for triumph, and we've had some big hits, predicting teams such as those previously mentioned Patriots in '06 and Cleveland State last year. This time around, we have further refined our approach. We have zeroed in on team stats that correlate strongly with upset wins and losses in past tournaments. We've conducted multiple regression analyses, which essentially is a way to tell how strongly each member of a group of inputs (those stats) affects an output (giant-killing success or failure). And we have devised a model to estimate each team's potential to be a successful Killer or Slain Giant and to specify the chances for an upset when two particular teams meet.
Through our GK blog, we will maintain and update our lists of potential killers and vulnerable giants every day. Each team is rated on a scale of 1 to 100 according to how closely its statistical profile resembles underdogs that have pulled off huge upsets in past NCAA tournaments or Goliaths that have gone down early. For now, these lists are purely relative. We're confident that Cornell has a much better chance of killing a Giant than Winthrop, based on its rating, but we won't offer actual odds of an upset until we see each team's opponent.
By definition, potential Giant Killers appear to be significantly worse teams than the Giants they play. So why do some win anyway? Well, luck is a huge factor. We haven't seen the spreadsheets yet that can explain why Vermont beat Syracuse in 2005; the Catamounts probably just had a great night while the Orange happened to play at the bottom of their range. But Giant Killers do share many similarities, because successful underdogs are often better than traditional basketball statistics make them look.
For one thing, college hoops teams play at a wide variety of tempos; Division I squads have ranged from 57.5 to 84.8 possessions per game this season. Pace determines raw point totals for individual teams, but efficiency, or points per possession, is the key to determining winners. Utah State, for example, scores 73.1 points per game, which doesn't look like anything special, but it plays at a glacial pace (333rd in the country). The Aggies actually score 115.5 points per 100 possessions, ranking 14th in the nation. Whether they force their opponent to play at their preferred or are compelled to play up-tempo basketball themselves, the Aggies remain efficient. So our model makes sure to adjust for pace.
For another, strength of schedule has an enormous effect on team stats. Favorite example: Coastal Carolina held opponents to less than 60 points on its way to 28 wins this season but padded its record with victories over Voorhees, Allen, Bridgewater (Va.) and Cornell (in Iowa, not New York). (The Chanticleers lost to Winthrop in the Big South championship.) And our research indicates schedule strength has a surprisingly strong impact on teams' upset chances, too. Our model accounts for this by factoring in the adjustments developed by Ken Pomeroy on his advanced metrics site.
But successful underdogs aren't simply hidden gems in terms of quality. Typically, they also employ strategies that increase the variability of their scoring, as this ESPN The Magazine piece explains. Long story short -- if you average 70 points per game, you have a better chance of pulling off an upset if you score 90 on some nights and 50 on others than if you constantly hover between 75 and 65. (You also have a better chance of getting blown out, but who cares?)
So it's no surprise to find that Giant Killers employ high-risk/high-reward strategies. Some try to generate extra possessions by pressing, going for steals, aiming for blocks or crashing the offensive boards. Others attempt to maximize the value of possessions by taking bunches of 3-point shots. Again, if you try these strategies and fail, you can easily find yourself down 20 points within a matter of minutes. But successful Killers execute. Statistically, they have:• Low turnover rates and high rates of generating opponent turnovers.
• High offensive-rebound percentages.
• High 3-point scoring as a proportion of all points scored.
As for the rest of our secret sauce? Well, we can't disclose all the ingredients. And we also can't predict every upset -- where would the fun be in that?
To see those teams most likely to slay giants and the behemoths most likely to fall, plus get access to all of our NCAA tournament content, you must be an ESPN Insider. You can sign up here.
MORE MEN'S COLLEGE BASKETBALL HEADLINES
- Duke's Hood follows Parker into NBA draft
- Silver: Making NBA age limit 20 top priority
- Brown becomes 4th Butler player to transfer
- Shot-blocker Obekpa returning to St. John's
MOST SENT STORIES ON ESPN.COM
ESPN INSIDER'S SECRET FORMULA
Giant Killer Stories
Keep tabs on all Giant Killers content here.
Giant Killers Blog Posts
Live updates before and during the tournament.