Mailbag: The past success of Giant Killers

Kenneth Faried & Co. gambled from deep against the Cards. Doug Pensinger/Getty Images

As full slates of NCAA tournament games finally get under way, we thought this would be a good time to put together some last-minute Giant Killer ideas and respond to questions from you, our loyal readers. On a few occasions, we've engaged in the debates raging over Giant Killers in the comments to our posts, drawing this classic response from Tom Butters of Toronto: "Hi Peter, thanks for the explanation. The column is still lame ... But thanks for explaining why." Now it's time to address your queries in more detail.

First things first: We exclude teams from our analysis because of the definitions of Giants and Killers, not because we dislike them. For those of you still wondering why Gonzaga is a Giant or whether Georgia is a potential Killer or what our statistical model thinks of Wofford, you can find links to our methodology and our evaluations of the matchups in every region of the tournament on right side of our Giant Killers blog.

Next, thanks where it is due -- to all of you who have offered feedback, whether through the GK Mailbag or in comments sections on our blog or elsewhere. We'd like to give a particular shout-out to Zachary Richardson of Denver, whose comments and questions, last year and this, prodded us to make significant improvements to the part of our statistical model that predicts head-to-head outcomes.

Now, into the mailbag for a few pressing questions:

Anonymous asks: "What is the breakdown of Win-Loss records for teams in each group of [Giant Killer success] percentages?"

Since 2004, potential Killers with GK scores of 50 or more have won five of seven, or 71. 4 percent, of games against Giants.

At scores from 40 to 49.9, GKs have won eight of 18, or 44.4 percent.

At scores from 30 to 39.8, GKs have won six of 14, or 42.9 percent.

At scores from 20 to 29.9, GKs have won six of 33, or 18.2 percent.

At scores from 10 to 19.9, GKs have won three of 37, or 8.1 percent.

At scores below 10, GKs have won one of 56, or 1.8 percent. (That one was Bucknell over Kansas in 2005.)

Ryan Ifland of Washington (we're not sure if that's the district or the state) writes: "I did some number crunching of my own on Belmont and found out that in their four losses, they average 27.8 percent from the 3-point line, minus-2 on the offensive glass, minus-9.5 on the defensive glass, minus-2 in the block column and 20.25 fewer free throw attempts. They do average plus-4 in turnover differential and shoot 41 percent, but I can't see them getting any further than the Sweet 16.

P.S. I haven't finished looking at the rest of their stats besides the four losses."

First of all, if Belmont makes the Sweet 16, that would be an enormously successful run for the Bruins.

But more important, be very careful about drawing conclusions from how bad potential Killers look in their losses. Remember, Giant Killers pursue high-risk/high-reward strategies, trying to increase the number and/or value of their possessions to boost their odds of beating better teams. When that works, they can pull off big wins. Morehead State not only got 17 rebounds out of Kenneth Faried, it hit 9 of 19 3-point shots to beat Louisville. But when it doesn't work, Killers can look as bad as Saint Mary's or Cornell did in the third round of the tournament last year -- or as bad as UAB did on Tuesday night. Greater variability means better odds of being on the wrong end of blowouts.

What's really problematic about Belmont's four losses is that three of them -- and zero of its wins -- came against the three teams that were by far the best opponents the Bruins faced all season.

Evan Schlossberg of Los Angeles writes: "I'm a huge college basketball fan that, I'm guessing like you, spends a lot of time with Excel open (unfortunately for me, the data I'm working with doesn't have columns for points and rebounds). I'm wondering how you calculate the metrics that you always reference, and [if there] is some sort of magical spreadsheet that is loaded up with stats for every D1 team and player. If so, how can I get access to this? I will pay you for it."

There is indeed a magical spreadsheet for Giant Killers, but its precise contents are top secret and held under heavy guard at the offices of ESPN The Magazine by day and under my pillow at night. However, if you want player stats for categories such as points and rebounds, you can get them from the NCAA. And if you really want to home brew, it's important to get familiar with and use tempo-adjusted metrics. You can find those at Ken Pomeroy's site, where nearly every number links to an interesting, sortable list.

Andrew of Toronto asks: "Why is it that if a potential GK is, say, a 10-seed, and it faces a giant that's a 7-seed, it's not eligible to be a Giant Killers matchup? I understand your larger point that to be considered a true Giant Killer, you need to beat a team ranked substantially higher. But two teams' relative strength and vulnerability shouldn't change if it's a 7-10 matchup as opposed to a 6-11, right?"

Fair point, Andrew. But we had to pick an upset threshold somewhere, and from the earliest days of this project, a gap of five seeds seemed like a much bigger deal than a gap of three. Historically, 10-seeds have beaten 7-seeds just over 40 percent of the time, while 11-6 upsets have occurred in 31.7 percent of matchups. It's subjective, but we think the tipping point that says "big upset" is between those two numbers, probably around 2-1 odds (33.3 percent).

Also, if Giant Killers were purely about picking upsets on their merits, we could rate every game where one team is measurably worse than another according to some objective scale, like Las Vegas point spreads. But GK is also a guide to the early rounds of your brackets, which means we aren't just trying to spot the squads most likely to beat better teams, but also to exploit irregularities and mistakes in the NCAA's seedings. And those are clearer in 6-11 and 5-12 matchups than 7-10s or 8-9s.

Finally, as a practical matter, there are no teams seeded seventh or 10th this year from non-power conferences. So at the moment, we're not leaving any potential matchups between Giants and Killers out of our analysis.

Finally, Tayler4747 says: "Thanks for making me look up which team was the Monarchs and which was the Blackbirds. It isn't exactly common knowledge, you idiot."

As if we all needed any more reason to root for Wofford and Boston University, here's hoping they both go on extended runs and meet in a Terriers versus Terriers national semifinal, just to keep things simple for Tayler!