Ducks Attack Message Board Forum

http://espn.go.com/college-football/bcs

Sagarin who actually usually has duck friendly rankings, is now a joke.. Baylor is ranked ahead of both Stanford and Clemson.

Now I don't understand why the BCS uses Sagarin's elo_chess rankings and not his normal rankings:
http://www.usatoday.com/sports/sagarin/fbt11.htm

Someone please elaborate and try to justify the Baylor ranking. Thanks.

Don't worry about these numbers until the end of the year when all the data has been collected.

UOducksTK1 wrote:http://espn.go.com/college-football/bcs

Sagarin who actually usually has duck friendly rankings, is now a joke.. Baylor is ranked ahead of both Stanford and Clemson.

Now I don't understand why the BCS uses Sagarin's elo_chess rankings and not his normal rankings:
http://www.usatoday.com/sports/sagarin/fbt11.htm

Someone please elaborate and try to justify the Baylor ranking. Thanks.

Human polls should be banned. We should be using Sagarin's Predictor column (and scrap everything else).

The elo_chess rankings use wins and losses only. The predictor rankings use points. The BCS banned using points in computer formulas years ago which is why he added the elo_chess version. Jeff Sagarin himself disagrees with the elo_chess ranking and has been an advocate for using points. It makes no sense not to.

Using points is tricky though because it encourages teams to run up the score and leave their starters in, both of which I'd rather not see incentivized.

That is very easy to address.

Sagarin's Predictor ranking uses a diminishing returns factor. So, once you're up by a safe amount, there is little ranking advantage to be gained by scoring more.

wheaton4prez wrote:

UOducksTK1 wrote:http://espn.go.com/college-football/bcs

Sagarin who actually usually has duck friendly rankings, is now a joke.. Baylor is ranked ahead of both Stanford and Clemson.

Now I don't understand why the BCS uses Sagarin's elo_chess rankings and not his normal rankings:
http://www.usatoday.com/sports/sagarin/fbt11.htm

Someone please elaborate and try to justify the Baylor ranking. Thanks.

Human polls should be banned. We should be using Sagarin's Predictor column (and scrap everything else).

The elo_chess rankings use wins and losses only. The predictor rankings use points. The BCS banned using points in computer formulas years ago which is why he added the elo_chess version. Jeff Sagarin himself disagrees with the elo_chess ranking and has been an advocate for using points. It makes no sense not to.

Ah I see.. Why shouldn't points be a factor? Perhaps a minor factor since teams like Houston will just run up scores for the purpose of boosting rankings, but I mean points should definitely play a role in top 25 matchups or against quality teams.

UOducksTK1 wrote: Ah I see.. Why shouldn't points be a factor? Perhaps a minor factor since teams like Houston will just run up scores for the purpose of boosting rankings, but I mean points should definitely play a role in top 25 matchups or against quality teams.

Like Wheaton said.... Diminishing return...

For example, a 14 point win would be the same as a 28 point win (or something like that)....

uodux1212 wrote:

UOducksTK1 wrote: Ah I see.. Why shouldn't points be a factor? Perhaps a minor factor since teams like Houston will just run up scores for the purpose of boosting rankings, but I mean points should definitely play a role in top 25 matchups or against quality teams.

Like Wheaton said.... Diminishing return...

For example, a 14 point win would be the same as a 28 point win (or something like that)....

Right. But does that term include throwing out non-competitive teams? Perhaps only games vs ranked teams should include points/margin of victory.

UOducksTK1 wrote:Ah I see.. Why shouldn't points be a factor? Perhaps a minor factor since teams like Houston will just run up scores for the purpose of boosting rankings, but I mean points should definitely play a role in top 25 matchups or against quality teams.

Originally, the computer formulas did use points and some coaches figured out that they could run up the scores to game them. So, instead of adding a cap or diminishing factor, the BCS completely banned points from being used.

They should be used. The actual scores are concrete data, based on the rules of the game, that are much more telling than a W or L.

UOducksTK1 wrote:Right. But does that term include throwing out non-competitive teams? Perhaps only games vs ranked teams should include points/margin of victory.

I don't think that would be possible, mathematically.

The points are used to establish ranking to begin with. So, how could it throw out unranked teams to establish ranking, before the ranking formula has completed processing the teams?

Also, the point margins in games against unranked opponents can be telling. If a ranked team beats a weak opponent by 1 point when every other team beat them by 20 points, that ranked team should take a hit.

Including margin of victory but with a diminishing return or "cap" probably does make some sense. But if, say, it is understood that a BCS computer has a hard cap of 35 points for a margin of victory (i.e., it does not benefit a team to win by more than that), you still have to figure that teams with a 28-point lead who have the ball with two minutes to go will be going all-out to get one more TD instead of running out the clock. Besides how that might "look," there's the added issue of keeping first-string players in longer, which might lead to more injuries that will hurt the team more in the long run. So if there is a way of having a diminishing value of margin of victory built into the computers but somehow keeping it a "secret" as to how it's done (if that's possible), maybe that's the way to go . . .

And I'll add that, IIRC, it was largely because of how the Ducks got shut out of the NC game after the 2001 season that the BCS threw out margin of victory as a factor in the computer rankings.

Greenblood wrote:Including margin of victory but with a diminishing return or "cap" probably does make some sense. But if, say, it is understood that a BCS computer has a hard cap of 35 points for a margin of victory (i.e., it does not benefit a team to win by more than that), you still have to figure that teams with a 28-point lead who have the ball with two minutes to go will be going all-out to get one more TD instead of running out the clock. Besides how that might "look," there's the added issue of keeping first-string players in longer, which might lead to more injuries that will hurt the team more in the long run. So if there is a way of having a diminishing value of margin of victory built into the computers but somehow keeping it a "secret" as to how it's done (if that's possible), maybe that's the way to go . . .

And I'll add that, IIRC, it was largely because of how the Ducks got shut out of the NC game after the 2001 season that the BCS threw out margin of victory as a factor in the computer rankings.

Im no math/statistical wiz by any stretch... but maybe they could lower the threshold? Use margin of victory up to 14 points and then anything after that is considered a "blow out"... The 14 point delta wouldn't change many end of the game plans in coaching decisions IMO.