| Rank | Team | POFF Prob | Change |
| 1 | Clemson | 100.0% | 0.9% |
| 2 | Alabama | 100.0% | 10.8% |
| 3 | Notre Dame | 100.0% | 0.0% |
| 4 | Oklahoma | 45.5% | 9.7% |
| 5 | Ohio State | 24.9% | 3.8% |
| 6 | Georgia | 18.2% | -26.0% |
| 7 | Michigan | 7.9% | 0.3% |
| 8 | Central Florida | 1.9% | 0.4% |
| 9 | Washington State | 1.7% | 0.1% |
Five years into the Playoff era, and my simple little model appears likely to have correctly "predicted" 19 of the 20 playoff participants (oops). I would pat myself on the back, but most of the selections were so obvious that I could make an even simpler model that would perform pretty much just as well.
The weird-at-the-bottom, boring-at-the-top 2018 season held to form, as the four teams likely to make the Playoff started the season 1st (Clemson), 2nd (Bama), 4th (Oklahoma), and 7th (ND) in my Playoff odds. We were essentially one weird night game at Purdue away from complete chalk, and thankfully my alma mater turned out to be the beneficiary.
As for where I'm at headed into year 6, I feel pretty good. The one thing that bugs me is that the Michigan-UCF-Wazzu triumvirate at the bottom should probably have smaller odds, especially because it's way less likely that a 2-loss team sneaks in with all of the undefeated teams in place. I have an idea that might mitigate this problem (basically adding in a chalkiness factor as another variable to my logit), but I'll have to play around with it in the offseason to see if it actually works.
No comments:
Post a Comment