I did it the least squares, mathletics way (whilst solver can still handle less than the 200 variables) – I haven’t added weightings to recent games and have an anomaly for Youngstown State and Cincinnati (due to the need to take all of the results in) and the model is definitely off (standard deviation is 7.99 rather than in the around 16 range). Home advantage is just over 2.89 and here is my top 20 teams through week 4:

Cincinnati 59.38958988
Youngstown State 55.11057165
Oregon 46.75626512
Delaware State 46.11646678
Texas 45.77080435
Alabama 44.25723436
Texas Tech 41.30113876
Pittsburgh 38.41650154
Texas A&M 35.58243228
Louisiana State 35.35818466
Florida 33.57554718
Georgia 33.33801369
Kansas State 32.53842928
Florida State 32.00777582
Oklahoma State 30.57184235
Stanford 30.07745255
Purdue 29.08145694
South Carolina 28.81613351
Arizona State 28.68139892
Fresno State 28.51357736

Points are all in relation to an average team (e.g. an Arkansas at 0.087) being rated 0 (e.g. so Oregon is 46.75 points better than average). I will try and add this in for next week but I expect there might be more than 200 variables involved by that point (so solver will cry enough!).

Nice post and great work as usual.

Purdue has been very consistent this year; they blew out two bad teams and played to an SRS draw (road loss of 3) against a very good team (at least so far):

Date      tm1     pf  tm2               pf  h/r  w/l diff  MOV   CON  CON   SOS   SRS
1-Sep-12  Purdue  48  Eastern Kentucky   6   H   W    42   31.5  B10  fcs   27.7  59.2
8-Sep-12  Purdue  17  Notre Dame        20   R   L    -3    0.0  B10  IND   55.6  55.6
15-Sep-12 Purdue  54  Eastern Michigan  16   H   W    38   29.5  B10  MAC   24.9  54.4