College Football’s Biggest Upsets
I’m back with more riveting statistical analysis for the world. Try to stay calm and please, do not mob me when I go out in public. Just be patient as I have never refused a photo or autograph request from a fan.
This newest list will once again be based on my ratings, which of course will introduce a few constraints on the analysis. My historical ratings are obviously year-end results and not even I am interested in recalculating each season by taking out the upset in question and then assigning a value. So this analysis is based on each team’s final rating for the full season (including the upset). The z-table was limited to one-hundredths accuracy, meaning that rating differences were rounded. This results in some of the games having the same percentage, but they are sorted by actual difference.
My first step was to take the Top 1000 teams in the All-Time ratings set and look at any losses they suffered. I then took the difference in the final ratings between the two teams, applied a standard z-table, and determined the percentage likelihood that the better team would win a rematch according to the final ratings. The effect of the constraint listed above now becomes apparent. Because the computer used the actual upset in its calculations, it sees the upset as a “true” evaluation of the two teams’ abilities. Therefore the percentages that this technique produces are deflated relative to if I had recalculated the year’s ratings without the upset and then looked at the matchup. Each team’s worst loss was analyzed and if that loss had a percentage of 80% or greater - remember, this is the probability that the ratings say the loser would win a rematch - then their next worst loss was looked at. This ensured that all losses scoring greater than 80% among the All-Time Top 1000 teams were accounted for.
So here are the 25 Biggest Upsets in College Football History using only the top 1000 teams and losses (ties not included). Season ranks are based on D-1A teams only.
| Rank | Winner (Season Ranking) | Loser (Season Ranking) | |
|---|---|---|---|
| 1 | 1981 Georgia Tech (102) | 1981 Alabama (9) | 98.26% |
| 2 | 1978 Memphis (77) | 1978 Houston (11) | 95.73% |
| 3 | 1980 New Mexico (86) | 1980 BYU (14) | 95.64% |
| 4 | 1982 Washington St. (71) | 1982 Washington (8) | 95.35% |
| 5 | 1999 Cincinnati (72) | 1999 Wisconsin (6) | 95.25% |
| 6 | 2007 Stanford (68) | 2007 Southern Cal (6) | 95.05% |
| 7 | 1977 Mississippi (44) | 1977 Notre Dame (3) | 94.74% |
| 8 | 1989 Southern Miss (46) | 1989 Florida St. (4) | 94.06% |
| 9 | 2007 Pittsburgh (61) | 2007 West Virginia (5) | 93.45% |
| 10 | 2001 Oklahoma St. (62) | 2001 Oklahoma (6) | 92.79% |
| 11 | 1961 TCU (32) | 1961 Texas (2) | 92.51% |
| 12 | 1998 Michigan St. (33) | 1998 Ohio St. (2) | 92.22% |
| 13 | 2002 Texas A&M (47) | 2002 Oklahoma (4) | 91.77% |
| 14 | 1991 Southern Cal (39) | 1991 Penn St. (6) | 91.62% |
| 15 | 1996 Memphis (57) | 1996 Tennessee (8) | 91.62% |
| 16 | 1976 Purdue (40) | 1976 Michigan (3) | 91.47% |
| 17 | 1993 Minnesota (50) | 1993 Wisconsin (8) | 91.47% |
| 18 | 1998 North Carolina St. (34) | 1998 Florida St. (3) | 91.31% |
| 19 | 2003 California (34) | 2003 Southern Cal (2) | 91.15% |
| 20 | 1976 Mississippi (39) | 1976 Georgia (4) | 91.15% |
| 21 | 1957 Kentucky (48) | 1957 Tennessee (8) | 90.99% |
| 22 | 1970 Texas A&M (59) | 1970 LSU (9) | 90.32% |
| 23 | 1978 Oklahoma St. (62) | 1978 Missouri (15) | 89.62% |
| 24 | 2001 Auburn (28) | 2001 Florida (2) | 89.44% |
| 25 | 1980 San Jose St. (62) | 1980 Baylor (15) | 89.25% |
At this point I always like to clarify that this list is not intended to be the final say in this discussion, only a starting point. Obviously some of the games are so close in their ranking that even this system doesn’t really declare a clear winner. Also, as I’m sure many of you are wondering, Bob Stoops also appears on the list at #28 (2007 Colorado) and #30 (2002 Oklahoma St.); Mack Brown has never lost an 80% game. Hey, at least it’s something, right? Pete Carroll also appears at #33 (2006 UCLA) and #72 (2006 Oregon St.). I’m starting to think that the more great teams you have the more likely they are to be upset.
One of the things I noticed was the lack of old games. Part of it is certainly parity, but I think the biggest factor in this discrepancy is that more games are played, giving the system a better idea of teams’ true strength. Back in the old days when a team would only play one or two games against good teams, those games are all the system has to rate the teams. So what was probably a huge upset (like Centre over Harvard in 1921) is not considered as big in my ratings because the winner is inflated due to the lack of other information.
After this was complete I decided to throw in some other games. To do so I used the following links as a starting point:
Biggest Point Spread Reversals
A blogger’s list of big upsets
Ivan Maisel’s biggest upsets written in 2002
I also decided to throw in ties because of Maisel’s article and the only big one I could think of was Texas A&M and SMU back in the day. So here is a table that includes all games in the above links plus that A&M/SMU tie and Appalachian St./Michigan game, followed by a table showing the biggest upsets that Texas has been involved in. However, I did leave out the games in the above links that are not considered upsets under this system.
| Rank | Winner (Season Ranking) | Loser (Season Ranking) | |
|---|---|---|---|
| 1 | 1985 UTEP (108) | 1985 BYU (22) | 99.68% |
| 2 | 1994 SMU (85) (GAME TIED) | 1994 Texas A&M (10) | 98.54% |
| 3 | 1998 Temple (93) | 1998 Virginia Tech (21) | 98.12% |
| 4 | 1980 Georgia Tech (83) (GAME TIED) | 1980 Notre Dame (8) | 97.19% |
| 5 | 1992 Iowa St. (69) | 1992 Nebraska (13) | 94.06% |
| 6 | 2007 Appalachian St. (64*) | 2007 Michigan (13) | 88.10% |
| 7 | 1986 California (52) | 1986 Stanford (15) | 86.65% |
| 8 | 1985 Oregon St. (77) | 1985 Washington (37) | 85.54% |
| 9 | 1942 Holy Cross (57) | 1942 Boston College (17) | 80.23% |
| 10 | 1972 Missouri (31) | 1972 Notre Dame (14) | 69.5% |
| 11 | 1949 Auburn (65) | 1949 Alabama (41) | 69.15% |
| 12 | 1969 San Jose St. (71) | 1969 Oregon (45) | 65.91% |
| 13 | 1921 Centre (27) | 1921 Harvard (20) | 56.36% |
* - Appalachian State ranking based on where they would have been if they were D-1A
| Rank | Winner (Season Ranking) | Loser (Season Ranking) | |
|---|---|---|---|
| 1 | 1961 TCU (32) | 1961 Texas (2) | 92.51% |
| 2 | 1941 Baylor (59) (GAME TIED) | 1941 Texas (3) | 89.07% |
| 3 | 1962 Rice (52) (GAME TIED) | 1962 Texas (8) | 88.01% |
| 4 | 1996 Oklahoma (78) | 1996 Texas (23) | 87.70% |
| 5 | 1999 Texas (23) | 1999 Nebraska (3) | 86.43% |
| 6 | 1968 Texas Tech (41) | 1968 Texas (3) | 86.21% |
| 7 | 1978 Baylor (39) | 1978 Texas (8) | 85.99% |
| 8 | 1948 Texas A&M (68) (GAME TIED) | 1948 Texas (17) | 84.61% |
| 9 | 1992 TCU (91) | 1992 Texas (48) | 83.15% |
| 10 | 1976 Texas (29) (GAME TIED) | 1976 Oklahoma (9) | 81.86% |
| 11 | 1994 Rice (72) | 1994 Texas (33) | 81.33% |
| 12 | 1965 Rice (71) | 1965 Texas (27) | 81.33% |
| 13 | 1967 Texas (33) | 1967 Oklahoma (2) | 81.06% |
| 14 | 1996 Texas (23) | 1996 Nebraska (5) | 80.51% |
| 15 | 1991 Texas (46) | 1991 Oklahoma (17) | 80.51% |
| 16 | 1980 Texas (29) | 1980 Oklahoma (6) | 80.23% |
| 17 | 1957 Texas (50) | 1957 Baylor (22) | 80.23% |
The 1968 Texas Tech loss is obviously a big one when viewed at the end of the season, but we know that the new system was still being implemented early in the year.
If any of you have some big upsets in mind that I’ve missed, let me know and I’ll plug in the numbers.
August 11, 2008 at 12:30 pm
I’m puzzled over the lack of old games. I wonder if it has something to do with the nature of the game. In the old days, there might be just a few scoring opportunities. Teams had their fastest players playing safety, to keep the long score from happening. Teams would punt on 3rd down to try and get a little better field position.
You would think that in such a game, an underdog could get lucky on just one play that makes the difference (as did happen to the Horns in ‘61). Maybe it just wasn’t that simple.
August 11, 2008 at 12:45 pm
I really think it has to do with the number of games and even moreso the number of interregional games.
Back in the old days there weren’t enough different paths by which to connect two teams. So their head-to-head result had a greater impact on their ratings relative to each other. Now, when Stanford beats Southern Cal, there are a large number of game paths that discount that result and “prove” that the Trojans are a better team than the Cardinal. The fewer games and interregional matchups there are, the less contradictory evidence available.
August 11, 2008 at 1:03 pm
What about the 1941 7-7 tie between Baylor and Texas?
Texas was 6-0 averaging over 38 points a game while only giving up 27 total points in the 6 contest. Baylor ended up the season 3-6-1, with the three wins over Hardin-Simmons, the University of Denver, and Arkansas.
Texas was #1 and was going to be on the cover of Life Magazine the next week. The 7-7 tie only dropped them to #2 in the rating, but the 14-7 loss to TCU (who played in the Sugar Bowl that year) dropped Texas to #4 in the final AP Poll.
August 11, 2008 at 1:30 pm
89.07% on that Baylor tie. So it would move into the #2 spot in Texas upset history.
August 11, 2008 at 1:34 pm
Nice breakdown. Out of sheer curiosity where would Texas ‘upset’ of USC rank? There are certain ESPN inboxes that need said information.
August 11, 2008 at 1:45 pm
N/A
Not an upset. Wasn’t even an upset before the game as Texas was rated #1 in almost every computer rating, including mine. Massey, Sagarin, etc.
Basically all the good systems.
Anyway, the number is 69.15% for the 2005 Texas/USC matchup. The system says Texas would win basically 7 out of 10 matchups.
August 11, 2008 at 2:12 pm
Agreed. Texas over USC was an upset of perception. Nothing more. We were a dominant team by any objective measure.
Thanks, Huck. This was cool stuff.
August 11, 2008 at 2:36 pm
“89.07% on that Baylor tie. So it would move into the #2 spot in Texas upset history.”
Thanks Huck. While you are at it, could you run the numbers on the 14-14 tie with Rice in 1962?
Rice won only two games that year, and the tie cost Texas the National Championship. (They were #1 at the time and didn’t lose until the Cotton Bowl to LSU).
I have always assumed that had Texas beaten Rice and been declared National Champs, we would have had the Final Poll vote changed to after bowl games a few years earlier that it happened.
August 11, 2008 at 3:14 pm
BC ultimately didn’t mind losing in 1942 to Holy Cross 55-12 (#9 in your second list), though it cost them the #1 ranking at the end of the season. Since they lost, the players cancelled their plans to celebrate at the Cocoanut Grove nightclub later that evening. 492 unfortunates attended, though, and were burnt to death when the place caught on fire.
http://en.wikipedia.org/wiki/Cocoanut_Grove_fire
August 11, 2008 at 3:16 pm
Where would 1942 Northwestern (1-9) 3, Texas (9-2) 0 rank?
August 11, 2008 at 3:44 pm
77.04% on the Northwestern loss. The war years have some weird ratings situations and I’m thinking about rerunning those. Northwestern played an incredibly difficult schedule that year, though, which I’m sure depresses the result.
The 1962 Rice/Texas tie is 88.1%.
The numbers in this post are approximations as the file isn’t here and it’s chicken scratch on a napkin type calculations.
August 11, 2008 at 3:57 pm
“The numbers in this post are approximations as the file isn’t here and it’s chicken scratch on a napkin type calculations.”
That’s okay, you have always been clutch when it comes to numbers.
August 12, 2008 at 5:30 am
Added the above games as well as some others to the Texas table.
August 12, 2008 at 7:55 am
How about Royal’s first win over OU, 15-14 in 1958.
August 12, 2008 at 8:00 am
My dad maintains that the biggest upset in history (and labeled “The Upset of The Decade” for some time) was Tech’s 1966 win over Arkansas to knock them out of the national title hunt. He reminds my mother about 4 times a year that she skipped the game to go Christmas shopping with her little Chi-O friends because Tech sucked that season and it was cold outside. Followed by a loud thrashing of a retro Saddle Tramp bell.
August 12, 2008 at 8:02 am
Huckleberry, could you see where that one ranks, it might have been 1968, not for sure? I would like to let my dad know exactly where it stands based on sufficient statistical analysis.
August 12, 2008 at 9:24 am
1958 Texas over Oklahoma is 70.88%. Seems like it should be higher to me as well.
I believe 1966 is the game you’re talking about, ded. It rates a 79.67%. Arkansas was 8-1 going into the game and Tech was 3-6.
August 12, 2008 at 9:36 am
That would be the one. I actually need to correct myself, it was labeled “The Upset of the Century” for a while. Looks like my dad is a hot, mouth liar. I feel like the dude in Big Fish.
August 12, 2008 at 9:49 am
It’s also funny that every time he tells the story, Tech’s record at the time gets worse and worse. I was convinced they were 0-36 through 9 games that year.
August 12, 2008 at 9:58 am
Well, Arkansas was definitely a top team then. They had been upset by Baylor, who finished 5-5, earlier in the year. But from 1964 until the bowl game after the 1965 season they had won 21 games in a row.
I’m sure that colors his memory of the game. Meanwhile, Tech had gone on a 5-game losing streak earlier in the year and the 3 teams they’d beaten finished with a combined total record of 8-20-1.
August 12, 2008 at 10:04 am
Has anybody else been catching any of those SWC Season Recaps from the 60s and 70s on Fox Sports SW? I saw the one for 1966 the other day and called my parents to make sure they caught. It featured the Tech/Ark game at the very end. There was/were car/cars parked on the sidelines. My dad said he used to drive that old thing to haul guests of honor around before the game. It was just weird seeing a guy streak down the sideline for a TD and then he passes a Cadillac.
August 18, 2008 at 6:17 am
Your second table includes a bunch of games that aren’t ties or D1AA games but nevertheless are not on the first list, even though you said the second list was everything on the first list plus the SMU-A&M tie and the Appalachian St-Michigan upset. What is the actual criteria for the 2nd list that adds the other 11 games? Losing team wasn’t considered top 1000?
August 18, 2008 at 6:20 am
No, the second list is all the games from the links I provided plus the extra couple of games. Games from the first list weren’t in that one.