As we have rolled out our Giant-Killer predictions for first-round upsets in the NCAA tournament, we've been getting a lot of queries that turn out to be different versions of the same question:
Why have we told people to stay away from Virginia Commonwealth versus UCLA when analysts across the country are picking Eric Maynor & Co. as the underdog du jour?
Why have we called North Dakota State a better bet against Kansas than Cleveland State against Wake Forest, when our own numbers seem to suggest otherwise?
Why do we have "unrivaled hatred" for Michigan State?
Here's why: It's all based on history. And history shows that safe Giants are better bets than dangerous Killers.
Our statistical analysis looks at teams before the 65 NCAA entrants are even selected. Thus, it makes a general estimate of teams' chances of winning, but it's not matchup-based. Sometimes, though, two teams that each have strong (or weak) profiles will meet, creating a potential quandary.
So let's peer deeper into the spreadsheets.
Suppose we sort Giants into two groups we'll call Vulnerable Giants, who have a 50 percent or greater chance of falling, and Safe Giants, whose odds are less than 50 percent. And that we do the same for Giant Killers, creating Effective Killers (50 percent or better chance of knocking off a Giant) and Ineffective Killers (less than 50 percent).
The killing of a Giant is by definition a big upset, which obviously limits the number of Vulnerable Giants and Effective Killers. And that means most tournament games involving potential Killers actually take place between Safe Giants and Ineffective Killers. As you would expect, Giants dominate these contests, going 57-9 (.864) from 2006 through 2008.
Also, as you might expect, Giant Killers do very well in the exact opposite situation, when Effective Killers face Vulnerable Giants. There have been only two such matchups in the past three tournaments, but Killers won both.
Here's where things get interesting. When Ineffective Killers have faced Vulnerable Giants over the same span they have won five of seven games (.714). But when Effective Killers have run into Safe Giants, they have won only three of five games (.600). These sample sizes are small, but the numbers suggest that when strength meets strength or weakness meets weakness, the Giant's vulnerability matters more.
Put another way, you can gamble on a scattershot Killer to pull a big upset, but you shouldn't bet against a powerhouse Giant. Since 2004, five of 21 successful Giant Killers (23.8 percent) had less than a 20 percent chance of winning, according to our model. But only two of 25 fallen Giants (8 percent) had a score of lower than 20 percent.
So, while we'd love the chance to pick VCU, given the fact that the Rams top our GK rankings, UCLA's 0 percent score makes that move unwise. In a vacuum, we'd love to pick Cleveland State, but Kansas has a better chance of losing to any Killer than Wake Forest does, which makes North Dakota State a better play. And we have nothing against Michigan State; it just counts for a lot that our model assigns the Spartans a 41.7 percent chance of falling.
Which brings us to one more lesson. Last year, a Vulnerable Giant played an Effective Killer twice (Drake versus Western Kentucky, Georgetown versus Davidson), resulting in two memorable but eminently predictable Giant slayings. And from 2006 through 2008, there were 13 NCAA matchups involving either a Vulnerable Giant or an Effective Killer -- that is, games where our analysis says at least one team had at least a 50 percent chance of contributing to a huge upset.
But in the first round of this year's tournament, there's exactly one: Xavier versus Portland State. The rest of the potentially Giant-Killing faceoffs you'll see on Thursday and Friday -- all 18 of them -- involve Safe Giants playing Ineffective Killers.
We're not complaining. But the more we ponder the first-round matchups, the more we're looking forward to the second round.
Peter Keating is a senior writer for ESPN The Magazine.