CSNbbs

Full Version: BCS computer anomolies
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I'm struggling to understand how the computers are so high on:

Kansas State - 7th in the computers
Baylor - 15th
Texas - 16th
Auburn - 21st

What is with the Big 12 love from the computers? Am I missing something? Those rankings are a travesty.

Meanwhile those getting screwed by the computers:

Wisconsin - 28th
Florida State - 29th

Wisconsin (and maybe Florida State) would beat the bejesus out of any of those teams listed above.
The computers aren't allowed to use margin of victory, which is a better determinant of how good a team is than simply winning and losing, which is all they are allowed to use. I forget when the BCS made the ratings developers take it out, but Jeff Sagarin ranted about it after they did it and basically said they were significantly reducing the quality of the ratings. He still publishes both because he can't stand the version he's forced to use for the BCS. The actual scores of games provides more data points than simply winning and losing, which is essentially 0 (loss) or 1 (win). The more data points, the more useful and granular ratings become.

I have ratings I created in college for a class that I still update and I can run them with margin of victory and without, and the one with margin of victory included is significantly more accurate in predicting results and aligns much more closely with perception. There are always some oddities because some teams play very difficult or very easy schedules and are actually much better (and in some cases, worse) than their record would indicate (Mizzou and Texas A&M are the poster-childs for that this year; very good teams trapped in a mediocre team's body because of their schedules) but, in general, they're much better than ratings that only take into account winning and losing.

Funny thing is, ratings with margin of victory also help non-BCS schools because there is no limit on how high they can climb, because they could theoretically win every game by 80 points. That eliminates the issue of their conferences being weak and, as a result, they can only climb so high before their schedule holds them back if the rating is limited to just winning and losing. You can't do anything about New Mexico being 1-9, but you CAN blow them out. The current BCS rules force the ratings systems to remove the "blowing out" part of the analysis.

The computers like the Big XII because, outside of Kansas, there are no other terrible teams. Everybody else is 5-5 or better.
(11-13-2011 11:03 PM)UIHuskie Wrote: [ -> ]The computers aren't allowed to use margin of victory, which is a better determinant of how good a team is than simply winning and losing, which is all they are allowed to use. I forget when the BCS made the ratings developers take it out, but Jeff Sagarin ranted about it after they did it and basically said they were significantly reducing the quality of the ratings. He still publishes both because he can't stand the version he's forced to use for the BCS. The actual scores of games provides more data points than simply winning and losing, which is essentially 0 (loss) or 1 (win). The more data points, the more useful and granular ratings become.

I have ratings I created in college for a class that I still update and I can run them with margin of victory and without, and the one with margin of victory included is significantly more accurate in predicting results and aligns much more closely with perception. There are always some oddities because some teams play very difficult or very easy schedules and are actually much better (and in some cases, worse) than their record would indicate (Mizzou and Texas A&M are the poster-childs for that this year; very good teams trapped in a mediocre team's body because of their schedules) but, in general, they're much better than ratings that only take into account winning and losing.

Funny thing is, ratings with margin of victory also help non-BCS schools because there is no limit on how high they can climb, because they could theoretically win every game by 80 points. That eliminates the issue of their conferences being weak and, as a result, they can only climb so high before their schedule holds them back if the rating is limited to just winning and losing. You can't do anything about New Mexico being 1-9, but you CAN blow them out. The current BCS rules force the ratings systems to remove the "blowing out" part of the analysis.

The computers like the Big XII because, outside of Kansas, there are no other terrible teams. Everybody else is 5-5 or better.

Nice post!
Reference URL's