First off, an apology: our normally pretty good system went a solid 0-3. It required Tennessee to get outgained by nearly 300 yards and win, along with West Virginia failing to convert in the red zone, but it still went 0-3. This week, we’ll get at least one upset correct. Maybe.
- 50-59% (2-0 on the season, expected record of 2.32-1.68): #23 TCU (at Arkansas) (54.4%), #6 USC (vs. #14 Stanford) (57.6%), #17 Louisville (at North Carolina) (59.6%), #24 Notre Dame (vs. #15 Georgia) (59.9%)
- 60-69% (2-0 on the season, 1.38-0.62): #2 Ohio State (vs. #5 Oklahoma) (67.9%), #3 Clemson (vs. #13 Auburn) (69.6%)
- 70-79% (2-1 on the season, 0.71-0.29): #20 Washington State (vs. Boise State) (70.8%)
- 80-89% (3-0 on the season, 0.81-0.19): #4 Penn State (vs. Pittsburgh) (81.0%)
- 90-99% (11-0 on the season, 9.80-0.20): every other game
Yet again, no 90+% game is worth discussing. There literally isn’t a single one of these teams with a model that projects them to win by less than 25 points. Not interesting at all! Conference play will even this out. Anyway, two key notes: the 50-69% grouping last week was expected to go 2.47-1.53; they went 4-0. A 70-79% team lost, but it was the last one I would have expected. The 50-69% teams will not go undefeated this week; I can promise this. There’s a 5.3% chance they go 6-0.
Alright, let’s pick some losers. The most surprising part of our work this week is the following: the models give the top four teams (#5 Oklahoma left out as they are playing #2 Ohio State) just a 38.1% chance of going 4-0 with 0.82 expected losses. I highly doubt #1 Alabama will drop to Fresno State, and while it did happen last year, Pittsburgh beating a top ten Penn State team two years in a row seems not right. So it’s between two 60-69% games: #2 Ohio State vs. #5 Oklahoma and #3 Clemson vs. #13 Auburn.
These are basically a coin flip, and I’m ready to lose on another toss-up, so I’m taking #13 Auburn over #3 Clemson. Why? Well, consider it two-fold: Auburn’s got a recruiting talent advantage over Clemson, and when removing the largest outlier model for both this game and OU-OSU, this is a slightly more likely event (66.9% vs. 69.9%). There’s a 53% chance that either #2 or #3 lose on Saturday night; good enough for me.
We still need two more upsets, and both are likely from the 50-59% range. We need one more from the Top 25 vs. Top 25 grouping. How about #14 Stanford over #6 USC? This knocks out our necessity of finding an upset in the Top 15. Again, after removing the outlier model, this one was a slightly more likely event than Georgia defeating Notre Dame. I also like this one because it hits on some concerns with USC: the significant schematic advantage Stanford’s offensive staff likely has over USC’s defensive staff, plus USC’s significant struggles with a Group of Five team in the first game. Perhaps USC’s defense ends up being pretty good, but I’m not yet sold.
For our final upset, we need one from the low end of the Top 25. It needs to be an unranked team. We’ve got two contenders: #17 Louisville losing at a mediocre North Carolina team or #23 TCU losing at a probably good-ish Arkansas team. You can probably guess which way I’m swinging here: Arkansas over #23 TCU. Arkansas, based on stats from CFB Matrix, has a large coaching advantage on offense versus defense over TCU. Plus, it’s a home game, the first CBS game of the year, and Arkansas has both a small recruiting talent advantage and doesn’t have to play the reigning Heisman Trophy winner.
To recap, our picks this week are #13 Auburn over #3 Clemson, #14 Stanford over #6 USC, and Arkansas over #23 TCU. Other games considered: #15 Georgia over #24 Notre Dame, North Carolina over #17 Louisville, #5 Oklahoma over #2 Ohio State, and Pittsburgh over #3 Penn State. Three upsets should come from this group of eight games. If there is a fourth upset, look for Boise State over #20 Washington State.
gyrateplus: the website 