Per Bill's request, here are the top 30 1-A upsets of 2005-2010, as calculated by Compu-Picks:
|1||Stanford||Southern California||6||2007||24 - 23||AWAY||0.73|
|2||Maryland||Clemson||5||2009||24 - 21||HOME||0.72|
|3||Indiana||Northwestern||9||2008||21 - 19||HOME||0.67|
|4||Pittsburgh||West Virginia||14||2007||13 - 9||AWAY||0.66|
|5||Miami (Ohio)||Bowling Green State||8||2008||27 - 20||AWAY||0.62|
|6||Oklahoma State||Texas Tech||11||2005||24 - 17||HOME||0.6|
|7||Syracuse||Louisville||4||2007||38 - 35||AWAY||0.59|
|8||Northwestern||Iowa||10||2009||17 - 10||AWAY||0.59|
|9||Iowa State||Missouri||12||2006||21 - 16||HOME||0.59|
|10||Washington State||Southern Methodist||3||2009||30 - 27||HOME||0.56|
|11||East Carolina||North Carolina State||7||2010||33 - 27||HOME||0.56|
|12||Southern Methodist||Texas Christian||2||2005||21 - 10||HOME||0.56|
|13||Nevada-Las Vegas||Utah||4||2007||27 - 0||HOME||0.56|
|14||Eastern Michigan||Bowling Green State||6||2008||24 - 21||AWAY||0.55|
|15||New Mexico State||Nevada||7||2008||48 - 45||AWAY||0.55|
|16||Mississippi||Florida||5||2008||31 - 30||AWAY||0.55|
|17||Purdue||Ohio State||7||2009||26 - 18||HOME||0.55|
|18||North Carolina State||Pittsburgh||4||2009||38 - 31||HOME||0.55|
|19||Virginia||Miami (Florida)||9||2010||24 - 19||HOME||0.53|
|20||Colorado||Oklahoma||5||2007||27 - 24||HOME||0.53|
|21||Rice||Southern Mississippi||6||2007||31 - 29||AWAY||0.53|
|22||Iowa State||Nebraska||8||2009||9 - 7||AWAY||0.53|
|23||North Carolina State||Boston College||4||2006||17 - 15||HOME||0.51|
|24||Texas-El Paso||Houston||5||2009||58 - 41||HOME||0.51|
|25||Kansas State||Texas||11||2006||45 - 42||HOME||0.5|
|26||Louisiana-Monroe||Troy State||10||2008||31 - 30||HOME||0.5|
|27||Nevada-Las Vegas||San Diego State||6||2005||13 - 10||HOME||0.49|
|28||New Mexico||Arizona||3||2008||36 - 28||HOME||0.48|
|29||Hawaii||Navy||13||2009||24 - 17||HOME||0.48|
|30||Syracuse||Notre Dame||13||2008||24 - 23||AWAY||0.48|
1) Of the top 10 upsets, all of them were 7 point or less wins. And of the top 25, only one (2007 UNLV over Utah) was a true beat-down. This is partially a reflection of the methodology (if you win by a lot, your rating goes up and the opponent rating goes down much more than if it's close), but I think that there's probably deeper meaning here.
2) Unsurprisingly, most of the upsets were at home, 19 of the top 30. Strangely enough, the very biggest upsets didn't seem to show much preference for home vs away (#1 upset was on the road and five of top ten). My guess is that this is just noise, and that with a larger sample size you'd see a smoother home/road distribution... but I could be wrong.
3) 2007 wasn't so much the "year of the upset" as it was the "year of the ROAD upset" (including the App St @ Michigan game not counted here) and "year of the MAJOR upset." 2007 had six of the top 30 upsets (nothing unusual there), but two of the top 4 (plus App St), and had four of the 11 road upsets on this list.
In terms of total number of shockers, 2008 and 2009 actually had 2007 beat, but their upsets tended to be of the non-headline variety, aside from Ole Miss over Florida, and to a lesser degree Northwestern over Iowa (though the Hawkeyes were practically begging to get tagged in 2009) and Purdue over Ohio St (though the Buckeyes had already lost at that point).
Additional notes / caveats:
1) The underlying ratings are based on the entire year’s results, NOT the expectations going into the game (i.e. TCU winning at Oklahoma in 2005 was not considered remarkable given how the two ended up, even though it was a big deal at the time). They are also NOT adjusted for HFA, byes, etc (obviously it’s easier to score an upset at home, but adjusting for that is a somewhat arbitrary process).
2) No special consideration was given to in-league, non-league, AQ vs non-AQ, etc. Interestingly, I found VERY few non-AQ vs AQ “major upset” type games. Mainly the meaningful upsets were intra-league play (including a few in the non-AQ group).
I suspect that this means that the non-AQ over AQ upsets which seem meaningful at the time turn out to reflect either that the non-AQ was better than people thought or the AQ worse (again, 2005 TCU over Oklahoma is a great example of this). It is (apparently) MUCH rarer to see a non-AQ who sucks throughout the year score one random upset over an AQ who is otherwise good throughout the year.
3) Again, 1-A vs 1-A only, since I don’t track AA games.
4) The underlying numbers are based on my own model, which I am currently tweaking as part of my annual adjustment process. Since the 2011 model has yet to be completed, these numbers will not necessarily reflect whatever process and/or calculation changes I have yet to make.