Latest Sagarin Rankings from 10/31

S

schuhbdoo

Starter
College Football 2010 through games of October 30 Saturday the BCS uses the ELO_CHESS from here

HOME ADVANTAGE= 3.03 RATING W L SCHEDL(RANK) VS top 10 | VS top 30 | ELO_CHESS | PREDICTOR

1 Oregon A = 94.54 8 0 71.48( 42) 1 0 | 3 0 | 93.43 2 | 95.25 1

2 TCU A = 91.57 9 0 67.47( 72) 0 0 | 2 0 | 90.43 5 | 92.31 2

3 Stanford A = 90.62 7 1 73.14( 25) 0 1 | 1 1 | 89.39 7 | 91.48 4

4 Missouri A = 90.16 7 1 74.89( 16) 1 1 | 2 1 | 90.82 3 | 88.92 6

5 Nebraska A = 89.51 7 1 72.53( 34) 1 0 | 2 0 | 90.82 4 | 87.71 9

6 Auburn A = 89.42 9 0 72.87( 29) 0 0 | 4 0 | 94.33 1 | 85.57 15

7 Boise State A = 89.28 7 0 66.98( 74) 0 0 | 2 0 | 87.53 13 | 90.89 5

8 Oklahoma A = 88.47 7 1 74.41( 18) 0 1 | 1 1 | 88.59 11 | 87.72 8

9 Alabama A = 87.72 7 1 71.96( 39) 0 0 | 1 1 | 84.58 17 | 91.97 3

10 Arizona A = 87.69 7 1 73.59( 21) 0 0 | 2 1 | 87.61 12 | 87.15 11

 
schuhbdoo said:
College Football 2010 through games of October 30 Saturday the BCS uses the ELO_CHESS from here

HOME ADVANTAGE= 3.03 RATING W L SCHEDL(RANK) VS top 10 | VS top 30 | ELO_CHESS | PREDICTOR

1 Oregon A = 94.54 8 0 71.48( 42) 1 0 | 3 0 | 93.43 2 | 95.25 1

2 TCU A = 91.57 9 0 67.47( 72) 0 0 | 2 0 | 90.43 5 | 92.31 2

3 Stanford A = 90.62 7 1 73.14( 25) 0 1 | 1 1 | 89.39 7 | 91.48 4

4 Missouri A = 90.16 7 1 74.89( 16) 1 1 | 2 1 | 90.82 3 | 88.92 6

5 Nebraska A = 89.51 7 1 72.53( 34) 1 0 | 2 0 | 90.82 4 | 87.71 9

6 Auburn A = 89.42 9 0 72.87( 29) 0 0 | 4 0 | 94.33 1 | 85.57 15

7 Boise State A = 89.28 7 0 66.98( 74) 0 0 | 2 0 | 87.53 13 | 90.89 5

8 Oklahoma A = 88.47 7 1 74.41( 18) 0 1 | 1 1 | 88.59 11 | 87.72 8

9 Alabama A = 87.72 7 1 71.96( 39) 0 0 | 1 1 | 84.58 17 | 91.97 3

10 Arizona A = 87.69 7 1 73.59( 21) 0 0 | 2 1 | 87.61 12 | 87.15 11
Click to expand...
This is due to our lower SOS. Sagarin takes that into account. Also our low score against SDSU hurts.

 
Yes, this is the drawback with computers. Sometimes you put the numbers in and the results don't make sense.

 
In Sagrin's defense, his ratings usually are pretty accurate at the end of the season. If you go to the actual site he offers a very detailed explanation of how the system works but a quick cover on it is this: The ratings are 50% based on wins and losses only, the Elo Chess rankings. Elo Chess are the rankings he turns into the BCS. The other 50% factors in margin of victory/defeat in with wins and losses, Pure Points, also known as the Predictor. Therefore the rankings you see there are the combination of both what you have done in the season, and how the computer thinks you will fare in the future. Losing to Texas by 7 and Missouri beating then #1 Oklahoma at the time by 9 is what is making their predictor score better than ours. Hopefully that help explains why we are behind them by, read his page for the official description.

 
Last edited by a moderator:
Yeah, Sagarin's model really is very well constructed. If you want to boil it down, basically Missouri is still edging us out in his rankings because despite the fact that we've had fairly similar SOS (theirs has been a little bit harder), and we both have a win over a top 10 team, their one loss came at the hands of the currently 5th ranked Nebraska Cornhuskers, while our one loss came at the hands of the currently 39th ranked Texas Longhorns.

 
Last edited by a moderator:
I read an interview with the late David Rothman a few years back and he said that head-to-head consideration was a fallacy that wasn't supported by any known theory of ratings. It kind of blew my mind, and I really didn't want to believe it because just about every fan, voter, talking head, etc. consider it when teams have equal records and have played each other. But I think it makes sense to me now. Our game with Missouri is counted, just not given any extra boost just because we happen to have the same record or because we are close to each other in the rankings. It is counted as one game, just like every other game, and we are given the appropriate boost in our rating because of having won the game.

Now, I am not completely decided on this and I'd like to hear from the other computer ranking owners on the issue, but... something to think about.

 
Last edited by a moderator:
Sagarin's rankings get better as the year progresses. It's rare you see these kinds of aberrations in his numbers, but they happen, and they're purely formula based.

One of the great things about Jeff Sagarin is, if you have a question and he's not too busy, he'll answer your email. I exchanged an email or two with him last year. Can't quite remember what it was about, but he was helpful and cordial, even though I'm an insect compared to his mathematic acumen.

 
sd said:
nuance said:
NU under Mizzou is a little wacky. But look at Auburn at #6. That's just plain crazy.
Click to expand...
the only argument is that our loss is a lot worse than their loss. still does not seem logical.
Click to expand...
I guess that’s the difference between human voters and code--or at least this code. Sagarin discounts head-to-head results as compared to a typical human voter. For two closely ranked teams, a human voter would always put the one who had beaten the other on top.

 
When you are ranking all 120 (or however many) teams, I don't see how you can consider head-to-head. Or, what you are really doing is considering head-to-head in every single game, which means that while our win over Missouri should seem to put us ahead of them, our loss to Texas yanked us below Missouri. It just doesn't make any season to look at all of the teams and the whole season to position all of the teams, and then say, "Oh yeah, plus I want to give more weight to that NU win over MU". Why would you do that?

Remember a few years ago when FSU was #2 with one loss, ahead of #3 Miami with one loss but Miami beat FSU? Many were in an uproar that this shouldn't have happened, how could FSU be ahead of Miami when Miami beat them?? But they wouldn't apply that same logic to #4 Washington, who had one loss, and beat Miami! Why wouldn't you look at that Washington-Miami game every bit as closely as Miami-FSU? In fact the system did, along with every other game in the season, and wound up putting FSU at #2.

 
The other thing is, computers can only use the information given to them. 8 or 9 games isn't a great sample to do a ranking on. 12 isn't either, but it's better. Humans can look at Auburn and say, I'll put them in the top 2 because I know that they'll have to get through the whole SEC by the end and they'll have earned that spot if they stay unbeaten. I think they have more of a tendency to look ahead and predict rather than swap unbeaten teams late in the year because one team's schedule got tougher as the season went on. The computers will swap teams without any conscience based on how it is told to evaluate the games that were played.

 
VA Husker Fan said:
When you are ranking all 120 (or however many) teams, I don't see how you can consider head-to-head. Or, what you are really doing is considering head-to-head in every single game, which means that while our win over Missouri should seem to put us ahead of them, our loss to Texas yanked us below Missouri. It just doesn't make any season to look at all of the teams and the whole season to position all of the teams, and then say, "Oh yeah, plus I want to give more weight to that NU win over MU". Why would you do that?
Click to expand...
If two teams end up adjacent in rankings and grade out within 1.5 percent, then it would make sense to put the team that just defeated the other on top--wouldn't it? I'm just saying that a human voter would probably always put NU over Mizzou if they graded out closely.

 
You must log in or register to reply here.
Back
Top