Sunday, March 16, 2008

Discussing "Clutch" vs. "Overrated"

One issue that I'm always juggling with and discussing on this website is the issue of teams that win a lot of close games. And I think that it's so important with respect to picking teams for brackets that I want to devote a long post to it:


An Analogy

Let's start the issue by dipping our toes into the shallow end of the pool, by discussing an analogy, the Pythagorean Expectation from Major League Baseball. It is a way of estimating a team's win-loss record by knowing just how many runs they've scored and given up. For the most part, every team in baseball finishes within five games of their "expected" value. The teams that finish above that value tend to be "clutch" teams. In baseball, the most likely teams to be "clutch" are those with great closers, so in recent years the most common team to finish with a better record than expected has been the Yankees.

What you see in baseball is that several teams get out of whack with their "expected" value, but they always return towards that value as the year goes on. It's not that teams that experience good luck in the early part of a season are destined to have bad luck in the second half, it's simply that random variables always tend to progress towards the mean over any given time. That's not to say that over an infinitely long season every single team would finish exactly at their expected value. It's simply that teams will finish close to that expected value, because as the sample size increases, the luck evens out.


What about basketball?

So let's go to basketball, where we all know that luck is often involved. That key shot that rims in or rims out? That debatable foul call? There are all sorts of games that are determined by luck. But at the same time, certain teams are more clutch than other teams. Typically teams that have strong senior leadership, and strong coaching. Also, teams that play a lot of close games tend to become better at close game situations over time, because the most important thing is experience. A final complication, of course, is that teams don't play the same schedules. So we can't rank teams purely on points scored/against like we do in baseball.

So what ranking do you suggest?

There are two good ones to look at: the Sagarin rankings and the Pomeroy rankings. The Pomeroy rankings try to break teams down by different efficiencies, and by taking out "luck", although I'm not entirely sure about how that is calculated. Overall, Pomeroy gives you a pretty good ranking system. So if you want to know which team is better in a given match up and don't want to just look at the seedings, check Pomeroy.

But the ranking that I'd rather discuss is Sagarin, because he splits teams by the ELO_CHESS and the PREDICTOR rankings. The former is a pure win/loss formula (similar to the RPI... only more complex and better), and the latter is heavily influenced by the scores of the games. If a team has a much better ELO_CHESS than PREDICTOR than they were either "clutch," "lucky," or some combination of the two. The ELO_CHESS more-or-less ranks teams like what you see in the bracket, while the PREDICTOR can often be pretty out of whack with the seedings.


So what do we make of a team with out-of-whack rankings?

Some people might interpret a team with a much better ELO_CHESS than PREDICTOR as a team you should pick in the Tournament because it's clutch. But I'd argue that this clutch-ness has already been built into their seed, which is based on games won & loss and not by the margin of those victories (for the most part). They've been so clutch so far, how can they be more clutch than they've already been? And more likely, like I said, teams will regress to the mean. So my interpretation is that the teams with inflated ELO_CHESS are over-rated, and these are teams you should be picking against.


So if you're right, who should I be betting on or against?

Here's an informal list of teams with very out-of-whack Sagarin rankings, and what to do if you agree with my analysis and don't take any other factors into account (factors that I'll get into later this week, such as home/road records). There are obviously other teams with more moderately out-of-whack rankings, which you can look through yourself. I'm just listing the most egregious examples among Tournament teams:

Likely to Over-Achieve:
Marquette
West Virginia
Kansas State

Likely to Under-Achieve:
Drake
Vanderbilt
Kent State

No comments: