Monday, March 15, 2010

Bracket Lesson #2: "Clutch" vs "Overrated"

I'm always fascinated when I hear the talking heads on ESPN complaining about how the Selection Committee is supposed to choose the 34 best teams for at-large bids. They're not, and they tell us that every year. They're selecting the 34 best resumes. There's a difference. And this is an opportunity for people filling out their brackets, knowing that there are steals to be had because some teams are seeded better than they would be if we were rating teams by how good they were.

A clutch team will win more close games than they lose, but even the most clutch team in the nation will lose close games when the sample size gets large enough. When teams have won a disproportionate percentage of their close games then they stop being clutch and they become lucky, and the lucky teams are the ones to bet against because their resume is not in tune with how good they are. There is a simple way to know when teams have a resume that is out of whack with their ability, which is to just look at the Sagarin ELO_CHESS and PREDICTOR. The Sagarin ELO_CHESS is by far the best objective projector of NCAA seed, while the PREDICTOR is (in my opinion) the best projector of future performance. Some believe that the Pomeroy rating is a slightly better projector of future performance, but overall they're close, and the Sagarin is more useful because it can be easily compared to the ELO_CHESS.

So we can expect teams with dramatically better PREDICTOR ratings than ELO_CHESS ratings to outperform their seed, and vice versa. One thing to remember is that this should not be the only metric that you use. As with all of these "bracket lesson" posts, they are just tools to help you with your bracket. I'll eventually preview all of the regions, but you can guess that if a team keeps showing up in posts of teams to bet on or against that it's a hint of where I'm going to tell you to go:

Teams that are better than their resume:
Georgia Tech
Utah State

Teams that are much worse than their resume:
New Mexico
Northern Iowa
New Mexico State

This year I do want to add in some Pomeroy ratings, because I know that for a lot of statisticians online they view the Pomeroy ratings as the best because they take into account all sorts of characteristics about teams rather than just scores of games. Pomeroy is a little bit more difficult to use because there isn't a resume rating, but we can use the Pomeroy Luck rating, which is a good gauge of how "lucky" a team has been (it's the deviation between the expected record and actual record). I don't view this as a metric as good as the Sagarin metric above, but you'll notice a whole lot of overlap between Pomeroy's lucky/unlucky teams and the Sagarin teams listed above:

Ten Worst Pomeroy Luck ratings:
330 - Houston
326 - Ohio
303 - Notre Dame
301 - Florida
299 - Marquette
297 - California
294 - Minnesota
267 - Georgia Tech
264 - Louisville
261 - Georgetown

Ten Best Pomeroy Luck ratings (ignoring 15 and 16 seeds):
1 - Oakland
3 - New Mexico State
18 - Tennessee
19 - New Mexico
20 - Kentucky
25 - Temple
34 - Gonzaga
38 - Cornell
39 - Sam Houston State
42 - Purdue

No comments: