Tuesday, October 05, 2010

2010 Gus Rankings: Week 5

Here are this week's Gus Rankings. Oklahoma tops the list, followed by TCU.

There was an interesting development in the Gus Rankings after this weekend's games, involving two of the more important games, that would happen only with great rarity in the traditional human-based polls. Florida lost to Alabama, yet remains ranked higher (3 vs. 4), and Stanford lost to Oregon, yet also remain ranked higher (14 vs. 18). How'd that happen? Florida's (and Stanford's) past opponents had a better weekend than Alabama's (and Oregon's). It's probably not likely to persist-- all four of these teams will probably win a vast majority of their remaining games, and these games will likely contain a large percentage of common opponents (with regards to the pairings above). So no large point advantage there. The advantage gained from last weekend's winners, however, is that they get additional points every time last weekend's loser wins another game. So in the long run, Gus pieces it out-- but the short-run effect is nonetheless interesting.

Art Carden requested a listing of the most overrated teams-- since that's a matter of relativity, I'll simply eyeball the largest differences (10 spots or greater) between the AP poll and the Gus Rankings, contingent on both teams being ranked. Feel free to compare yourself; I'll split it into two categories:

Gus likes, humans dislike:

- Virginia Tech (19 spots)
- Kansas State (15)
- Oklahoma State (13)
- Michigan (12)
- Missouri (12)
- Northwestern (12)
- Florida (11)
- Baylor (10)

Humans like, Gus dislikes:

- Arkansas (34 spots)
- Miami (FL) (25)
- Utah (22)
- Wisconsin (22)
- Oregon State (16)
- Oregon (15)
- USC (15)
- Nebraska (14)
- Ohio State (14)

It's interesting to look at-- there's 4 Big 12 teams on the "Gus likes" list, and 3 Pac-10 teams on the "Humans like" list, and Utah's joining that conference next season. Virginia Tech's difference may well be explained by the fact that Gus does not track games against FCS teams.

Actually, now that I've finished the list, I'd love to keep these groupings of teams in mind and see which group shakes out better over the course of the season.

4 comments:

siredge said...

I think people use rankings for two major purposes that are very different: a summary of past performance and a prediction of future performance. It seems clear that Gus rankings are primarily for summarizing past performance, but what adjustments would be necessary to make it useful for predicting future performance? One suggestion I heard was to adjust for major injury (ie Michigan if their QB gets injured). Any other thoughts?

Matt E. Ryan said...

Gus is actually fairly proficient at picking games; using the Sagarin rankings as a control, Gus reaches approximately that level by about Week 7 or Week 8, and (if I remember correctly) outperformed Sagarin during the bowl season.

Any objective ranking system, however, can only use past information to predict future performance. Issues of what to adjust for and exactly how to adjust introduce some subjectivity to the process.

siredge said...

Have you considered adding the conference best, average, median, and worst rankings each week?

siredge said...

Can I persuade you to post the Week 6 rankings? I'm hoping to use them to spice up a boring work training I've got to give at noon mountain.