Commentary

The return of Giant Killers

The secret to selecting NCAA tournament upsets is revealed below

Updated: March 10, 2011, 9:34 AM ET
By Peter Keating | ESPN The Magazine

Danero ThomasJed Jacobsohn/Getty ImagesRemember the Racers? Just one of many Giant Killer success stories.

It's become a March ritual as hotly anticipated as cherry blossom festivals, citizen lobbying days and Joe Lunardi dominating the Ocho: Welcome back to Giant Killers.

This is the sixth edition of our annual attempt to unlock the secrets of the opening rounds of the NCAA tournament, which happens to be the craziest, most unpredictable event in all of sports. If you are trying to decide if Jimmer Fredette or Nolan Smith should be player of the year, or whether Kansas has the chops to take on Ohio State if the schools meet in the national championship, you'll have to consult the many other fine analysts here at Insider. But if you want to know whether Morehead State has a prayer of surviving the first week of March Madness, or if there are any low seeds capable of knocking off Duke before the Blue Devils hit the Sweet 16, you've come to the right place. Here at Giant Killers Central, our season begins with the Big South tournament championship and usually ends before the Elite Eight. In between, we scrub mountains of data to uncover the huge upsets lurking in your brackets.

First, as always, some definitions. A Giant Killer is a team that beats an NCAA tournament opponent seeded at least five spots higher in any round. Squads from the six BCS power conferences are ineligible. We also exclude Butler, Gonzaga, Temple and Xavier -- nationally-ranked schools with lots of recent success, who aren't sneaking up on anybody. (Note that mathematically, any team seeded fifth or higher in its bracket, as San Diego State will be this year, cannot be a Giant Killer.) A Slain Giant, on the other hand, is simply a team that loses to a Giant Killer.

So last year, when Saint Mary's, a No. 10 seed, beat Richmond, a No. 7 seed, in the first round of the NCAA tournament, that was not a Giant Killing. But when the Gaels proceeded to knock off Villanova, a No. 2 seed, in the second round, that was a GK victory.

Since 2004, the first year for which we have collected data, 29 Killers have beaten 34 Giants. (A handful have slain multiple victims, as Cornell did in 2010.) What do these most successful deep underdogs have in common? Most fundamentally, they go high risk/high reward, thereby increasing the variability of their scoring, as we have explained in the Mag here and here. After all, if you average 70 points a game, you have a better chance of beating Goliaths if you score 90 on some nights and 50 on others than if you're constantly putting up between 68 and 72. (You'll also have a better chance of getting blown out, but this is March Madness, where a loss is a loss, so who cares?)

Every year, some Giant Killers attempt to generate extra possessions by pressing, going for steals, aiming for blocks and crashing the boards. Others try to maximize the value of their possessions by relying heavily on 3-point shots. A few, like Murray State in 2010, do both.

Remember, we said high risk/high reward. Try these strategies and fail, and you can be down by double digits in a matter of minutes. But successful Killers execute. Our statistical analysis indicates they have:

• High offensive-rebounding percentages
• Low turnover rates and high rates of opponent turnovers
• High 3-point scoring as a proportion of all points scored

Our research has also turned up several factors that make Killers better than traditional statistics make them appear. Strength of schedule is one. For example, North Texas (probable Sun Belt champion) and St. Peter's (MAAC champion) are both likely to be very low tournament seeds this year. But while the Peacocks have faced tough opponents such as Alabama and Seton Hall, the Mean Green have padded their record with the likes of Henderson State and Oklahoma Panhandle State (both unrated). History says the former approach sets teams up far better to pull off upsets once March Madness begins.

[+] EnlargeRyan Wittman
AP Photo/Kevin RivoliThe slipper fit Ryan Wittman and Cornell last season.

Further, college teams play at a wide range of tempos: Division I teams are averaging anywhere from 58 to 77.5 possessions per game this year, and these differences hugely distort box score stats. For instance, LIU, playing at the fourth-fastest pace in the country (74.2 possessions per game), has scored 80 or more 23 times this season. Old Dominion, playing at among the slowest (61.4 possessions per game), has broken 80 just three times all year. But the Monarchs are actually a more efficient offensive squad, scoring 108.9 points per 100 possessions to the Blackbirds' 107.5. And when two teams with different styles meet, it's efficiency, not raw totals, that determines the winner.

We have zeroed in on these and other factors that correlate strongly with upset wins and losses from previous tournaments, including strength-of-schedule data developed by Ken Pomeroy at his advanced metrics site. We have conducted multiple regression analysis, which is basically a statistical way to see how strongly each member of a group of inputs affects an output (in this case, the chances of Giant Killing success). And we have devised a model to estimate every 2011 team's potential to be a successful Killer or a Slain Giant.

In the coming days, we will provide and update lists of potential Killers and vulnerable Giants, and look at what's likely to happen when seeded teams collide. For now, we just want to note that if we aim our newly fine-tuned model backward in time and aggregate its predicted odds, it says there should have been 28.9 Giant Killing upsets since 2004. There have actually been 29. And it named the following teams as the most likely Giant Killers in each year: UAB in 2004, Wisconsin-Milwaukee in 2005, Bradley in 2006, UNLV in 2007, Davidson in 2008, Cleveland State in 2009 and Old Dominion in 2010. Every one of those teams did go on to slay a Giant.

We still can't explain why Kansas lost to Bucknell in '05. Luck is a huge factor in upsets. But to understand the part of the story that can be told, stay tuned.

Peter Keating is a senior writer at ESPN the Magazine, where he has covered investigative and financial stories since 1999. He coordinates The Mag's annual "Ultimate Standings" project, which ranks all pro franchises according to how much they give back to fans. He is the lead writer for ESPN Insider's annual Giant Killers project.

Peter Keating is a senior writer at ESPN The Magazine, where he covers investigative and statistical subjects. He started writing "The Biz," a column looking at sports business from the fan's point of view, in 1999. He also coordinates the Magazine's annual "Ultimate Standings" project, which ranks all pro franchises according to how much they give back to fans. His work on concussions in football has earned awards from the Deadline Club, the New York Press Club and the Center for the Study of Sport in Society.