Tuesday, September 22, 2009

2009 Gus Rankings: Week 3

Week 3 of the Gus Rankings are out, you can find them here.

A few things of note:

- Again, points are the wins of your opponents you defeated less the losses of the teams that beat you, concerning only games with two I-A teams.

Washington, which currently sits also at the top, is a good team to look at. They have 4 points because they defeated two teams-- USC and Idaho-- that have themselves each defeated two other teams. Further, the team that Washington lost to (LSU) hasn't lost any games yet, so they lose nothing for that loss.

Consider the possibilities for the rest of the season. If USC and LSU have great seasons, this further helps Washington. If they slip up and have less than phenomenal seasons, this mitigates the value of those wins. Win value now is not the same as win value later; this is different than the traditional polls, and I think it's an advantage.

Which bring us to the Idaho victory that is really helping Washington at the moment--I would gather that many people feel they are not too strong. What does this mean? As the season winds on, they will end up with a lukewarm record-- maybe somewhere around .500, maybe even less. This reduces the value of the Idaho victory for Washington as the season moves on; again, values of wins and losses change from week to week.

- I'm debating a few of the tiebreakers. Namely, if two teams have the same number of points, should a 2-1 team be ranked ahead of a 1-0 team? As it sits now, they would be-- though their winning percentage is less, and that tends to be a default tiebreaker in many cases. I'm noting how the measures do in predicting games (see next bullet point), so it's of importance now...but moving towards the future, as the complexity of the points system emerges, this will subsequently become less important.

- The benchmark for the Gus Rankings are the Sagarin rankings, the go-to power rankings (in my opinon) in college football. Taking last week's rankings and simply seeing if the higher team won (again, only in games in which both teams were I-A), the Gus Rankings were 25-22 and Sagarin went 38-14. Sagarin is able to predict a few more games than us because he has no ties in his rankings. We had quite a few ties last week but, again, that will diminish as the season goes on (13 different ranking levels last week, 32 this week).

- Southern Methodist went from #8 last week to #81 this week. Yikes!


Will Luther said...

It is really cool to see an objective measure of strength of schedule generating results where a 1 loss team ranks higher than undefeated teams. (Of course, it is still early and it remains to be seen how likely it is to generate such a result after all games are played, though this seems to be the where the conversation turns near the end of every season.)

I cannot help myself from imagining a more complex algorithm such that not only the teams you play are factored in, but also the teams they play, and the teams they play play, ...to the point where every team that is connected through the web of schedules is accounted for in each team's ranking. This would obviously be difficult to calculate by hand. You should venture over into the programming department to see if someone can set this up.

Ellis said...

How does the college Pythagorean predictor work?