Thursday, August 23, 2007

Non-archaeological post ESPN has an article up on preseason college football rankings and how predictive they've been for the last decade:
Past national champs among most overrated
The formula was simple: We took a team's preseason ranking in the AP Top 25 poll and compared it to where that team finished the season. For example, if State U. was ranked No. 6 in the preseason poll and finished 11th, the team lost five points. Conversely, if State Tech was ranked 13th in the preseason and finished second, it gained 11 points.

Any team that started or finished the season unranked was given a ranking of No. 26. For example, if State U. started the season ranked No. 7 and finished unranked, it lost 19 points, and vice versa.

Only teams that were ranked in at least three of the 10 preseason polls or final polls from 1997-2006 were included in the list of overrated and underrated teams. All teams included in at least one poll were included in conference standings.

The results of the survey are probably surprising to even the most die-hard college football fans. . .


I've been meaning to do something like this. I don't see why it should be all that surprising: traditionally high-profile programs will almost always be overrated because of their past, and vice-versa for relative unknowns. I've always thought preseason polls were largely a popularity contest anyway and ought not to be done until at least 2-3 games have been played.

There are some interesting things in the data though. WA State was probably bumped up so high due to the Mike Price years, where they performed much better than in the past. Same with Oregon (and Oregon State) who used to royally suck. Wisconsin was also traditionally weak until the mid-90s, and even then they tended to be underrated because they didn't generally have exciting offenses.

I don't get Washington though. The last few years they haven't been ranked and have performed as such. I thought they generally over-performed during the Neuheisel years, too. Be nice to see the raw data. . . .