Giant Killing 2.0: An updated guide

A look inside the numbers reveals that C.J. McCollum and Lehigh were a strong GK candidate in 2012. Streeter Lecka/Getty Images

The Ides of March are soon upon us, and you know what that means: It's time to sort the teams that are truly capable of pulling NCAA upsets from the tournament pretenders. As far back as the Rome Regional in 44 B.C., Julius Caesar sensed that "Yon Cassius has a lean and hungry look," according to veteran beat writer Bill Shakespeare. But J.C. misevaluated Brutus, and fell prey to a classic giant-killing. Too bad he didn't have our spreadsheets.

Here at Giant Killer Central, we have been isolating the keys to Cinderella success and boosting your early round brackets longer -- and dare we say better -- than just about anybody. We are proud to have predicted success for a series of low-seeded, unheralded teams that went on to slay Goliaths, from Cleveland State in 2009 to Cornell, Murray State and Old Dominion in 2010 to VCU and Richmond in 2011 to VCU (again) and Ohio last year.

But as this eighth edition of Giant Killers approached, we decided to make two major changes to our project.

For one thing, we have streamlined our definitions. In 2013, a "Giant Killer" is a team that beats an NCAA tournament opponent seeded at least five spots higher in any round. That's it, no exceptions. And a "Slain Giant" is simply a team that loses to a Giant Killer.

Previously, we had limited our definition of Giant Killers to non-BCS squads, not including a few highly successful schools such as Butler and Memphis. But with conferences constantly realigning and various mid-major programs rising in strength (and sometimes falling), these distinctions and exemptions stopped making sense. So we tossed them out. If Saint Louis gains a No. 6 seed and Iowa State a No. 11 seed in the same region this year, their meeting will be a Giant Killer matchup, with the Billikens as the Giant and the Cyclones as the Killer. (Mathematically, a team seeded fifth or higher in its bracket, as say Gonzaga and New Mexico are likely to be this season, cannot be a Giant Killer.)

We also wanted to get a better handle on just how much of Giant slaying is due to seeding, and what proportion springs from the factors that leads teams to play over (or under) their heads against tournament competition. So, with the help of GK contributor Neil Paine, we completely broke down our statistical model and reassembled it into a shiny new beast that hums with greater precision.

We started by examining all NCAA games since 2006, crunching that data into strength ratings for every team that has been to the Big Dance since then, and using those ratings to project results for each tournament game. This approach allows us to say things like, "Last year, Lehigh was 10.6 per 100 possessions points better than the average NCAA team after adjusting for strength of schedule, making the Mountain Hawks the best 15-seed since 2006, as teams seeded 15th are typically just 2.8 points better than average. That made an upset against Duke considerably more likely than most experts (and our old model) realized."

All that, however, was basically just our first step. We then moved on to ask: Comparing our projected results with what actually happened in the NCAA tournament, what explains the differences? Our new GK model isolates the statistical characteristics shared by Killers who surpassed expectations (and by bloodied Giants), so we can see which teams carry those same traits in 2013. Which means we can not only say that VCU is a very good team this season (23.4 points per 100 possessions better than average), but that the Rams have more in common with past Killers than any team in the country, boosting VCU's chances of beating a generic Giant to a whopping 83.4 percent as of March 4.

For the most part, our updated findings confirm what we've been writing for years: Giant Killers succeed by playing a high-risk, high-reward game, which widens the variability of their scoring. They crash the boards for offensive rebounds. They press, generating steals. They take a lot of 3-point shots. These strategies maximize the number and/or value of a team's possessions, improving its puncher's chance of beating a better opponent if everything goes well. True, they also increase a team's chances of getting blown out if things don't go well. But as we like to say here at GK Central, when it's win or go home, who cares if you lose by four points or 40 points? (Dean Oliver first laid out why inconsistency helps underdogs in his pioneering book, "Basketball on Paper," and you can see the reasoning laid out in colorful Bell curves at the bottom of this ESPN The Magazine piece.)

Conversely, Giants ward off assassins when they avoid sloppiness, play their normal game and allow quality to prevail. It's highly valuable, for example, for Giants to collect offensive rebounds, giving them second chances instead of allowing pesky Killers to build scoring runs. Our revised model adds a few new twists to describing this prophylaxis. It says, for instance, that the stronger a Giant's conference, the better it will swat away potential Killers come tournament time -- but that overall strength of schedule correlates negatively with Giant safety. We interpret this to mean that Giants benefit by banging against high-quality opponents most of the season, but also gain preparation from facing the kinds of lesser teams that can turn into Killers once March arrives. Maybe Louisville would have been better off in 2011 if it had faced a team like Morehead State in the regular season.

For those of you who like to play along with our statistical analysis at home, note that reverse-engineering the recipe for what we call the Giant Killers' "secret sauce" isn't simply a matter of finding the teams with the greatest variance in scoring. Many successful 12- or 13-seeds will spend most of the season dominating opponents from small or mid-major conferences, not necessarily engaging in high-risk tactics or changing their scoring patterns much, and then turn around and beat a much stronger team in the tournament. So we're hunting the ability to increase variance, which is a more elusive beast.

But we've tracked it, whether recent March Madness seasons were marked by lots of assassins (2010), heavier on vulnerable Goliaths (2012) or cloudy with chalk (2011). And we will follow the slingshots with even more accuracy this season. In the coming weeks, we will look at the key factors in big upsets, from pace to steals to the role of offense versus defense, as we analyze players, teams and matchups. Stick with us, and together we'll earn the appraisal Caesar made of the men who plotted giant killings in his day: "He thinks too much; such men are dangerous."