Sunday, March 14, 2010

Bracket Lesson #1: Which Upsets To Pick

I'll have a few bracket lessons that I'll go through tonight and through the next couple of days. These are all tested methods that have worked very well in the past, and will work well again this season. And the first lesson is, which seeds should you be picking as upsets, and I'll break this down into a few different sub-rules:

A) Don't pick 13, 14, 15 or 16 seeds to win a game, but keep those seeds in mind later: Even if you think the upset will happen, don't pick it. But keep that information in the forefront of your mind while picking future rounds. For example, if you think Purdue vs Texas A&M is an even match-up, you should consider taking Texas A&M just because you think there's a good chance that Purdue will lose in the first round.

B) Knock yourself out picking 9 & 10 seeds: Picking the 9 seed isn't even an upset, as the 9 seed has won 54% of the time over the 25 years of the modern 64/65 team bracket. And 10 seeds win their game 39% of the time, and since you're probably going to pick them to lose in the second round anyway you might as well take a shot on some 10 seeds.

C) 12 seeds win more often than 11 seeds: In the 25 years of the bracket, 12 seeds have won 34% of games against 5 seeds, while 11 seeds have only won 31% of games against 6 seeds. This seems counter-intuitive, but it's not if you read this blog regularly. I always talk about team psychology, and upsets are all about the favorite looking past their opponent. Teams that are 6 seeds know that their 11 was an at-large quality team, and they're hoping just to win one Tournament game. But 5 seeds tend to view 12 seeds as not even really deserving to be in the Tournament, and they're far, far more likely than 6 seeds to be looking ahead to things like Sweet 16 opponents.

D) Pick major conference 11/12 seeds to win, and mid-major 5/6 seeds to lose: I'm going to go lazy on you here and link to the analysis I did on this from two years ago (here). That analysis links to a big number crunch I did on 5/12 and 6/11 games back in 2006 (here and here, respectively). Some of the teams I talk about are a little outdated, but the general theme is apparent. Major conference 12 seeds actually win more than 50% of their games against 5 seeds. In fact, over the four years I've been advising people to take major conference 12 seeds, those teams are 4-2 in their first round games. Unfortunately, one of the best pieces of advice I have for you this season is useless, as we have zero BCS conference 12 seeds. But the other part of that advice was for you to avoid 5 seeds from mid-major conferences, and over the four years I've been giving that advice we've seen 5 seeds from mid-major conferences go 1-3 in their first round games. That makes Butler and Temple the obvious first round upset picks, although the one mid-major from the last 4 years to win a game as a 5 seed was Butler, over Old Dominion back in 2007.

Going over to the 6/11 games, the same advice holds, even though the data isn't quite as strong. But we've got examples this season, with Minnesota and Washington getting 11 seeds, and Xavier getting a 6 seed. With Minnesota and Xavier playing each other, you have give serious consideration to taking Minnesota as a first round upset pick.


Sandy Prisant said...

Just a small quibble about your list of teams that are better than their resume.
1. Minnesota is NOT one of the 34 best teams in America. It solely got in by having on its resume wins over all the Top 5 in the Big Ten. The Gophers ARE about resume over quality.
2. Utah State. The whole point of this school is to blatantly schedule no games to test their quality and always go for gaudy statistics in a weak conference. This is the most blatant example of resume over quality. By choice, this team has only played one Tournament calibre team (BYU) all season.

Jeff said...

I'm going to disagree on both of those teams:

Utah State: They did not just play one tournament quality team this year, they played several. They played Saint Mary's and lost by 5 points. They also went at Northeastern and lost by only 3 (in other words, if that game was played in Logan then Utah State probably would have won). And the WAC has some good teams in it: New Mexico State got a 12 seed this year, and Nevada also would have been a 12 if they'd won the WAC. And Louisiana Tech was a team I had on the bubble for much of the season. Against those teams Utah State went a combined 6-2, with both losses coming in true road games, and one of those coming right down to the buzzer. And the point of "resume over quality" is this: Utah State went 1-4 in games decided by five points or less, and 26-3 in games that were decided by six or more. So their resume was an unlucky resume, and should have been even better than it was.

Minnesota: Minnesota's big wins were the reason that they got in, but again they could have had a much better resume if a few balls had bounced another way. They had three losses out of conference, and they were against Texas A&M, Miami (Fl) and Portland. All three losses were by five points or less. If those three losses had been three wins, along with the impressive nine point win over Butler? That would have been a very impressive out of conference resume to go with their 12-10 Big Ten record (including tournament games). Compare that to a Virginia Tech team that went 10-7 against ACC teams, and played absolutely nobody out of conference. The only reason it was a debate between Minnesota and Virginia Tech was because of those three out of conference losses.

And again, this is not the first year I've given this advice. Go back and look at the last couple of years and look at the teams I've listed in the "Clutch vs Lucky" post. Basically every team that was "better" than their resume overperformed their seed, while basically every team that was "worse" than their resume went out in the first round. It's a remarkably accurate metric, and the best objective metric I'm aware of.