Georgia’s Second Standard, Air Force’s Burden of Proof: What the First CFP Rankings Showed Us

We got our first College Football Playoff rankings of the season last night, and there were few surprises. The five undefeated Power Five teams were the first five on the list. Alabama, Mississippi, Penn State, Oregon, Oklahoma, and Texas were the next six. After them it got a little funky—it was surprising, by precedent, to see Missouri and Louisville ahead of LSU and Notre Dame—and then it got weirder from there (Kansas State has a legitimate grievance), but it was mostly understandable, and the wackier parts will likely either end up being inconsequential or come out in the wash. Again, Kansas State has a legitimate grievance, and Georgia received unique treatment (we’ll get to that in a moment), but overall, it’s a fine set of rankings. There is nothing eye-popping that could affect the playoff.

Our college football model has a variable named FPA which we adjust based on each set of rankings the committee releases. FPA stands for “Forgiveness/Punishment Adjustment,” and it measures how far the committee is deviating from precedent established at the end of prior seasons. There’s no rule that each committee must be beholden to the work of its predecessors, so this isn’t some gradebook for the committee. Rather, it’s a measure of how surprised our model is to see different teams ranked in different places. Here’s how much FPA each team on the list got, in order of ranking:

Here’s that same list, in order of FPA, from positive (this team received the benefit of the doubt) to negative (this team did not receive any benefit of the doubt):

We talked about Liberty yesterday—Conference USA is so bad this year that it’s hard to guess how the committee views it, and as we suspected, we didn’t adequately outfit our model to handle this wrinkle. Put it on the list for next year.

We mentioned Georgia above: It’s notable that the committee put Georgia ahead of Michigan. Given neither has accomplished much so far this year, it signals that the committee thinks Georgia is a better team than Michigan. This is understandable—I’d guess a lot of people think Georgia is better than Michigan, and I personally kind of guess Georgia will be better than Michigan by the time it matters. But it implies that some mix of previous seasons and prior expectations and the recency of results play a role, and none of those are things the committee is supposed to weigh. Especially not the first two. The recency thing is more accepted (for reasons I don’t understand, games at the end of the season can sometimes matter more than games at the beginning of the season), but what the committee appears to have said here is that Georgia playing two good games (not in a row, mind you) is more important than Michigan playing five good games (in a row). Georgia is being graded on a different rubric, and it’s not going to matter—if they win the SEC or go 12–1, they’re almost 100% in, otherwise they’re almost 100% out—but it’s very interesting. Georgia has earned a lot. Have they earned the right to face a different set of criteria than the rest of the country?

Oklahoma State and Kansas State are interesting as well, and I think what’s happening here is the same thing that happened with Notre Dame and Louisville: The committee tried to avoid instances where they could be criticized for putting one team above another when the other has the head-to-head victory (and, in the case of ND and Louisville, a better overall record). If Kansas State loses to Texas, this doesn’t matter. If Kansas State beats Texas, this should be corrected, but we’ll be very curious to see to what extent. Kansas State was, through Week 9, our model’s 13^th-likeliest playoff team. That only came with a 1-in-70 chance of making the field, but our model now views it as 1-in-1,000. Kansas State has a legitimate grievance, and we’ll be curious to see how much it matters.

The thing of greatest consequence here is Tulane being ranked ahead of Air Force, because it signals that if each wins out, there’s a good chance Tulane retains that position. Most likely, this won’t be a concern—it’s unlikely either wins out, let alone both—but this is a big surprise and could affect which Group of Five team gets a New Year’s Six berth. The explanation is likely some combination of Tulane managing to schedule Mississippi and Tulane missing Michael Pratt in that game. It’s fair, but it’s a little surprising.

We’ve updated our bowl projections, both to account for these rankings and to account for last night’s games. The first big change: Texas’s average final ranking is now slightly higher than that of Notre Dame, pushing the Longhorns into a New Year’s Six bowl and Notre Dame out. The ripple effects of this are large and change just about every Power Five bowl matchup by sliding all Big 12 teams up a slot and pushing Notre Dame into the ACC mix, something which ultimately bumps Big Ten teams all down a slot by removing the Big Ten from the ReliaQuest Bowl. (It is really unfortunate that the ReliaQuest Bowl is no longer named the Outback Bowl or even the ReliaQuest Hall of Fame Bowl. We knew the quality of the Outback Bowl, and Hall of Fame Bowl would give some consistency as sponsors change. We cannot instinctively distinguish the ReliaQuest Bowl from the Guaranteed Rate Bowl or the 68 Ventures Bowl.) The second: Central Michigan is now likelier than not to be bowl eligible. We had Appalachian State in our projections yesterday, as the likeliest of the teams below 50% in bowl probability, but they’re now out of the mix.

For archival purposes, our bowl projections are below, as well as our latest FBS probabilities. We’ll have the full Week 10 preview tomorrow, along with a little update on the MAC after these Tuesday and Wednesday night games.