Here's my thinking: the transitive rankings I've been posting all season have results reasonably close to the computer rankings used in the BCS formula. Therefore, if we examine the effect of certain games on transitive rankings, we can surmise the contribution of those games on the official rankings.
In particular, we can figured out why the Big Ten has been poorly ranked overall by the computers, especially after the top three teams.
Before going game by game, a few overall revelations from the data: first, conference strength is about every team in the conference, not just the top. Second, the most revealing games are usually between teams with similar records in their respective conferences i.e. 7-1 Wisconsin beating 4-5 ASU isn't really a good measure and doesn't help the conference. Third, the MAC really sucks this year (even worse than usual).
Fourth, and most importantly, there simply isn't much basis to compare conferences. The eleven Big Ten teams played a combined 3 games against the Pac-10, 2 against the SEC, 2 against the Big 12, and 1 each against the Big East and ACC. This low connectivity causes these games generally to have a huge effect.
Note: 4 games against FCS teams were completely worthless. No one benefited from Indiana beating 1-10 FCS Towson.
For each game, it lists the national ranking of both teams and the total number of transitive wins or losses the game was worth to the conference.
#25 Michigan (7-5) beat #31 Notre Dame (7-5) 94
Michigan turned out to be a middle-lower Big Ten team, while Notre Dame had a number of quality wins (6 wins over bowl eligible teams). The net result was a whopping 94 transitive wins for the conference, or more than 8 per team (though they were not evenly distributed).
#51 Northwestern (7-5) beat #85 Rice (4-8) 79
Rice ended up going a better-than-expected 3-5 in CUSA, which really helped Northwestern (sadly, this win is the second best of the season for Northwestern). The Wildcats in turned "shared" this win with 5 other Big Ten members, so overall it helped the conference a fair bit.
#28 Iowa (7-5) beat #40 Iowa State (5-7) 73
Iowa State ended up 3-5 in the Big 12, but Iowa only ended up 4-4 in conference. To the computers, that's a decent win for the conference, especially since the Big 12 simply did not have very many out of conference losses. Running a similar analysis for the Big 12 says that this game is the worst loss for that conference.
#25 Michigan (7-5) beat #72 Connecticut (8-4) 62
Michigan beat the Big East conference champion. Unfortunately, transitive rankings (and most computers) really dislike the Big East, so it doesn't help as much as it might otherwise.
#18 Michigan State (11-1) beat #31 Notre Dame (7-5) 46
Like before, beating ND is a decent win. However, the 7-1 co-champion is expected to win this sort of game.
#8 Ohio State (11-1) beat #54 Miami (Fla.) (7-5) 40
In retrospect, Miami winning this game would have been a major upset.
#51 Northwestern (7-5) beat #90 Vanderbilt (2-10) 38
Vanderbilt is obviously not a great SEC team, but a win still helps a little bit. Of the wins over BCS teams, only Wisconsin over ASU was less impressive (due to Wisconsin's conference performance and ASU's FCS games).
#79 Indiana (5-7) beat #106 Arkansas State (4-8) 37
This game seems like it shouldn't be this high in the ratings, but it's Indiana's second best win and an (in conference) 1-7 team beating a 4-4 team always helps a bit.
#74 Minnesota (3-9) beat #112 Middle Tennessee (6-6) 34
When a team only has 3 wins total, every one of them helps quite a bit.
#39 Penn State (7-5) beat #68 Temple (8-4) 31
#61 Illinois (6-6) beat #69 Northern Illinois (10-3) 26
The MAC was just unbelievably horrible this year. Two 4-4 Big Ten teams beat the the two highest rated MAC teams (transitively speaking), but that didn't help the conference much overall.
The next 19 games all fall under "eating cupcakes" with the exception of Wisconsin beating ASU. However, two of ASU's wins are over FCS teams which don't help at all. Added to this is the "Big Ten Co-Champion should beat 4-5 Pac-10 team" factor, which results in the ASU game's inclusion with the others. Other than Wisconsin beating 1-11 UNLV and 1-11 SJSU, the rest of them are all wins over MAC, Sun Belt and FCS teams.
#8 Ohio State (11-1) beat #90 Marshall (5-7) 21
#25 Michigan (7-5) beat #109 Bowling Green (2-10) 19
#18 Michigan State (11-1) beat #99 Florida Atlantic (4-8) 19
#9 Wisconsin (11-1) beat #34 Arizona State (6-6) 18
#39 Penn State (7-5) beat #97 Kent State (5-7) 17
#8 Ohio State (11-1) beat #86 Ohio (8-4) 13
#79 Indiana (5-7) beat #117 Western Kentucky (2-10) 13
#84 Purdue (4-8) beat #109 Ball State (4-8) 12
#51 Northwestern (7-5) beat #119 Central Michigan (3-9) 12
#79 Indiana (5-7) beat #116 Akron (1-11) 10
#9 Wisconsin (11-1) beat #105 UNLV (1-11) 10
#25 Michigan (7-5) beat Massachusetts (6-5) 9
#18 Michigan State (11-1) beat #101 Western Michigan (6-6) 8
#28 Iowa (7-5) beat #109 Ball State (4-8) 8
#84 Purdue (4-8) beat Western Illinois (8-5) 7
#9 Wisconsin (11-1) beat #112 San Jose State (1-12) 4
#8 Ohio State (11-1) beat #119 Eastern Michigan (2-10) 3
#51 Northwestern (7-5) beat Illinois State (6-5) 3
#61 Illinois (6-6) beat Southern Illinois (5-6) 1
Now the losses:
#69 Northern Illinois (10-3) beat #74 Minnesota (3-9) -36
As previously stated, the MAC was horrible this year. Even a loss by a 2-6 Big Ten team to a division champ hurts quite a bit.
South Dakota (4-7) beat #74 Minnesota (3-9) -37
This loss should be much worse. I'm convinced that one of the reasons that transitive rankings have the bottom half of the Big Ten higher rated than most computer rankings is a mishandling of this game. North Dakota went 3-7 against FCS teams and 1-0 against Minnesota.
#28 Southern California (8-5) beat #74 Minnesota (3-9) -38
#13 Missouri (10-2) beat #61 Illinois (6-6) -43
#31 Notre Dame (7-5) beat #84 Purdue (4-8) -44
All of these losses were against teams that performed much better than their vanquished Big Ten foe. However, every loss still hurts.
#7 Alabama (9-3) beat #39 Penn State (7-5) -55
5-3 SEC team beats a 4-4 Big Ten team doesn't look good for the relative strength of the Big Ten.
#71 Toledo (8-4) beat #84 Purdue (4-8) -57
Losing to MAC opponents is never a good idea. Losing to a MAC team that lost to Wyoming is even worse.
#58 Fresno State (8-3) beat #61 Illinois (6-6) -60
Losing to a 5-3 WAC team doesn't look good either, especially since Fresno State lost to a dismal Ole Miss team as well.
#26 Arizona (7-5) beat #28 Iowa (7-5) -61
Arizona went 4-5 in the Pac-10. Iowa went 4-4 in the Big Ten. There isn't a better direct comparison of the two conferences.
One issue with this ranking is that is counts which games led to transitive wins, rather than doing some sort of "if this game didn't happen, what would be the net effect" analysis. The current method probably slightly overvalues wins over unique conferences (e.g. Northwestern was the only Big Ten team to beat a CUSA team, so most of the transitive wins for the Big Ten over CUSA teams go through the Northwestern-Rice game, even though most teams would have gotten those wins some other longer path had Northwestern not played Rice).
The final transitive win/loss records (not including Army-Navy, but that shouldn't have much if any effect):
|National Rank||Team||Transitive Win/Loss Record|