For the record, I'm thrilled that Football Outsiders is writing about college football. It seemed inevitable that they would eventually do so. The game is too big for FO to pass on all of those potential clicks. There was a dearth of good number-crunching sites for college football, so it's cool to now have FO on the job. Their major contribution to date has been to advance the theory that it makes sense when making predictions to look at a team's past five years as opposed to only one season. FO starts with a drive-based statistical measure (FEI) and then weights projected FEI based on a five-year average, along with returning starters and a conference adjustment. (I feel like I'm in second grade again in Ms. Watkins class delivering a book report.) This approach jibes with
something that I've been thinking, but can't quantify with numbers nearly as well as FO can: the simplest way to spot overrated teams is to pick against the nouveau riche teams in the top ten.
The upshot of this methodology in the SEC is that FO is
high on Auburn and down on Ole Miss($). This stat is scary for Ole Miss:
Since 2003, seventeen teams have registered a one-year improvement of at least six wins. Only three improved again the following season. Of the five BCS-conference teams to do the deed, all regressed by at least two wins the next year. The two most recent examples -- Illinois and Kansas in 2007 -- both regressed by four wins in 2008.
The caveat attached to it is reassuring:
Lucky for the Rebels, there are no clauses in the FO projections for "a potential Heisman darkhorse (Snead) transfers in to play quarterback" or "former head coach Ed Orgeron may have been historically bad at his job." Because of that alone, they may have a better chance than most at bucking what is a pretty solid statistical trend.
But screw their caveat; I'm still higher on Arkansas this year.
No comments:
Post a Comment