The Sunday Splurge is a Nerd

All Hail Non-Fuzzy Math

The last two weekends have illustrated why margin-of-victory is a relevant consideration. Take Sagarin's Predictor, for instance. The Predictor takes MOV into account. It was much higher on TCU than Utah because TCU doesn't have close calls against teams like Pitt and Air Force. The Predictor was very high on Stanford and the Cardinal dominated Arizona, the other challenger to Oregon in the Pac Ten. The Predictor was not high on Michigan State because of their multiple close games and the Spartans were duly exposed by Iowa. The Predictor was much higher on Arkansas than South Carolina and that played out on the field. Although the Predictor was reasonably high on Oklahoma, the Sooners' poor performance in College Station was presaged by their repeated narrow escapes earlier in the season.

Has anyone ever done a study comparing the predictive power of: (1) the human polls; (2) computer polls that do not account for MOV; and (3) computer polls that do. I know what the answer would be, but it would be nice to see a major outlet lay out the numbers in a mathematically sound manner. (The New York Times' filleting of the computer rankings is a good starting point.) There's a lot of self-interest in the media because most of the candidates to write such a piece are voters themselves and would not want to admit the fact that human polls are of limited value. That said, college football puts outsized importance on rankings, so someone ought to lay out the definitive case that we are using the wrong data in ranking our apples and oranges. On a related note: I'm genuinely interested as to the percentage of opposition to the BCS because of anger at the very idea of a two-team playoff as opposed to the ham-handed way in which the enterprise is managed (using inferior ranking methods, guaranteeing spots in BCS games for weak conference champions, etc.).  Put another way, would people like the BCS more if it were run by technocrats instead of boobs.

Their Epitaph Plain: Not Enough Offense in their Game

In July, I posited the following question about preseason #1 Alabama:

Has Saban ever coached a team with a "ridiculous" offense? The only one that comes to mind is his first SEC Championship team, the 2001 LSU team that survived an appalling pass defense to win the conference and the Sugar Bowl on the strength of Rohan Davey throwing to Josh Reed. Just as the 2007 LSU team was the weakest in recent memory to win a national title, the 2001 LSU team was one of the weakest in recent memory to win the SEC (unless you think that losing 44-15 at home to Florida is a sign of strength). Since 2001, Saban's teams have come off of an assembly line: functional offenses with quarterbacks who don't make mistakes, a rotation of athletic running backs pounding away, and a superlative defense highlighted by defensive backs who demonstrate the fruits out outstanding coaching. If the defense is a notch below that (and I'm not expecting TOO much of a regression), then the Bama offense will have to improve for the team to be better than 10-2. Can that happen? Jim Tressel and Urban Meyer met in the 2006 national title game with teams that played against their coaches' types. I'm fascinated to know if Saban can do the same.

We now how our answer.  Alabama has lost two games and will not make Atlanta without a bizarre succession of events involving losses by Auburn and LSU.  In the two games that the Tide lost, they scored 21 points each time.  Bama ran for 138 yards on 60 carries in the two games.  (Admittedly, sacks color this number.  It drives me crazy that college stats count sacks against the rushing game as opposed to the passing game.  This is one area in which college should mimic the NFL.)  Shouldn’t a team with two first-round picks at tailback and a bunch of returning starters on the offensive line do better than this?  Nick Saban is an outstanding coach in all sorts of ways, but he failed the test of being able to forge a championship team that is headlined by the offense instead of the defense.  Or maybe he just needs a better offensive coordinator.

How Much Old News Should we Read?

We’re headed for a new debate in deciding who goes to Glendale (assuming that at least one of Oregon or Auburn lose): should we account for evidence from previous seasons.  Boise State supporters are immediately going to latch onto their victory over TCU in Glendale.  Normally, this argument wouldn’t carry a lot of weight, but with the Broncos being almost entirely the same as last year (at least personnel-wise) and TCU also being fairly similar, the argument isn’t totally meritless.  SEC fans shouldn’t be dismissive of the argument, as one of the first arguments that we would make in favor of Auburn or LSU going to Glendale would be that SEC teams have won the last four national titles and are unbeaten in BCS Championship Games. 

This may be the one instance in which taking a prior year’s results into account makes sense.  We’re always dealing with a shortage of data in in-season college football debates, so why not, right?  (And I say this as someone who likes TCU more than Boise State and would prefer to see the Horned Frogs in Glendale.)  Knowing the BCS, Boise State will make the title game in part because of last year’s Fiesta Bowl, the Broncos will lose, and then the BCS will pass some ridiculous rule that will make the formula even shoddier than it is right now.