The fact that MOV isn't considered by the computers is really hurting BYU.

[jay santos]

Originally Posted by Cali Coug

http://www.usatoday.com/sports/sagarin/fbt08.htm

If MOV is considered, Sagarin has us at 11. When it isn't considered, we are at 35 (which gives us 0 BCS points for that poll).

Odd, given Bronco's aversion to running up the score. Even with that aversion, BYU is performing at a rate that would improve their ranking on Sagarin by 24 spots.

It's not that simple. The primary difference between his Elo Chess and the regular MOV ratings is that the MOV still has preseason bias.

For the first few weeks of the season, the starting ratings have weight
in the process(BAYESIAN), but once the teams are all CONNECTED, then the starting ratings are no longer used and all teams are started equaland the ELO-CHESS is then done in an UNBIASED manner from that point on.
RATING and PREDICTOR are now bayesian and the ELO-CHESS is now UNBIASED.

Computer models use a combination of game outcome vs SOS. Game outcome is usually a combination of winning percentage and MOV. But sometimes it is just winning percentage.

In winning perentage BYU is tied for #1 with 9 other teams. In MOV, rankings vary but my ranking on MOV with a diminishing principle has BYU at #5.

BYU vs Utah in MOV is not that dramatic to make a major difference. The main issue is the SOS.

[pelagius]

Originally Posted by jay santos

It's not that simple. The primary difference between his Elo Chess and the regular MOV ratings is that the MOV still has preseason bias.

I wouldn't go that far. The primary difference really is one uses MOV and the other doesn't. The Bayesian empirical adjustment for last season still matters but it is getting smaller in terms of its effect on the rating system.

I reproduced that Sagarin predictor without using an empirical bayes adjustment and there are some differences. Also it does look like the adjustment is helping BYU a bit. BYU is 17th in the non-adjusted MOV model and 11th with the empirical Bayes adjustment. My calculation of the Sagarin MOV (unadjusted) rating could be slightly off, but I can usually almost exactly reproduce Sagarin since he uses raw MOV in his predictor model (his predictor model is pretty trivial to reverse engineer).


Below should be Sagarin MOV top 25 without the Bayesian adjustment:

NCAA Football: 2008
Computer Rating Model: Margin of Victory (RAW)
Only Division 1A and 1AA games used

Rank  Team                    Rating   W   L
1     Southern Cal           118.478   4   1
2     Penn State             116.235   7   0
3     Texas                  115.361   6   0
4     Oklahoma               112.094   5   1
5     Florida                108.739   5   1
6     Missouri               108.549   5   1
7     Oklahoma St            104.538   6   0
8     Iowa                   101.333   4   3
9     California             101.152   4   1
10    Oregon St              101.103   3   3
11    Alabama                101.087   5   0
12    Texas Tech             101.050   6   0
13    Georgia                 99.937   5   1
14    TCU                     99.874   6   1
15    Ball St                 99.761   6   0
16    Michigan St             98.522   6   1
17    Brigham Young           97.058   6   0
18    Wake Forest             96.602   4   1
19    Boise St                96.478   5   0
20    Utah                    96.251   7   0
21    Arizona                 95.388   4   2
22    Tulsa                   94.826   6   0
23    North Carolina          94.819   5   1
24    Kansas                  94.775   5   1
25    Florida St              94.623   4   1

[Flystripper]

MOV used to be in the BCS computers. They took it out SU, not for ethics or sportsmanship, but because they believed MOV had too much weight since it is heavily considered in the human polls which have 2/3 weight.

[Indy Coug]

Originally Posted by Flystripper

MOV used to be in the BCS computers. They took it out SU, not for ethics or sportsmanship, but because they believed MOV had too much weight since it is heavily considered in the human polls which have 2/3 weight.

That is atrocious rationale. The sole purpose of any poll, computer or human, should be to project who would be expected to win in a given matchup. You cannot credibly model future outcomes of football games without MOV in that model.

[Flystripper]

Originally Posted by Indy Coug

That is atrocious rationale. The sole purpose of any poll, computer or human, should be to project who would be expected to win in a given matchup. You cannot credibly model future outcomes of football games without MOV in that model.

not my rationale

[jay santos]

Originally Posted by Flystripper

MOV used to be in the BCS computers. They took it out SU, not for ethics or sportsmanship, but because they believed MOV had too much weight since it is heavily considered in the human polls which have 2/3 weight.

I don't think this is correct. I think they took it out for ethics reasons.

[pelagius]

Originally Posted by jay santos

I don't think this is correct. I think they took it out for ethics reasons.

Here a quote from an ESPN article at the time. It seems to suggest you are right (but maybe the article is just short on nuance):

The changes pleased the American Football Coaches Association, which had pushed for such adjustments since February 2000.

"It's exactly what we asked for," said Grant Teaff, the association's executive director, which represents all college football coaches. "The big thing for the coaches was certainly the margin of victory. We didn't think it was right and the coaches didn't want a system where the more you score, the higher you rank."

[Indy Coug]

Originally Posted by Flystripper

not my rationale

I understand the rationale is not yours.

[BlueK]

Originally Posted by SeattleUte

OSU is of note. It has the inside track on the Pac 10 title, and beat USC. Michigan and AFA are not of no note, either.

please, is this a joke? Navy and Toledo must be national championship contenders. There you go again proving you don't know jack squat about sports.

[Mars]

Couldn't a system that uses MOV add some sort of "score cap" feature?

Meaning that Arizona's 70-0 win over Idaho doesn't count as anything more than a 21-0 win, but Auburn's 3-2 win over Miss State still only counts as a 1-point victory...