Yesterday’s CFP selection show was the most surprising in the twelve-year history of the playoff. Ohio State jumping the 2014 Big 12’s co-champions was newsworthy, but it wasn’t as surprising as yesterday. Florida State missing the 2023 playoff at 13–0 was controversial, but it wasn’t as surprising as yesterday.
When there are surprises with these things, there’s a lot to learn. Here’s what we learned, and what we’ll try to build into our model going forward into next season:
1. SEC Championship Losses Don’t Count.
We now have eight data points on teams who lost power conference championship games in the twelve-team era. So far, the Big 12 and ACC championship losers have been passed by one or two idle teams each, while the SEC and Big Ten championship losers have been passed by zero idle teams. All the ACC and Big Ten championship losers lost close. Both the Big 12 championship losers were blown out. One SEC championship loser was blown out; one lost close.
What to make of all of this?
With one caveat (we’ll get to that in #3), we won’t expect teams who lose the SEC Championship to drop in rankings going forward. If Alabama couldn’t drop with Saturday’s loss, we don’t know how anybody could. For the Big Ten, ACC, and Big 12, the picture’s murkier. It might depend on whether a loss could eliminate the team in question from the playoff field. Going forward, we’ll adjust our playoff selection model to take a more nuanced approach to this situation, while acknowledging the inherent randomness in this area of the process.
2. Head-to-Head Matters a Lot.
Notre Dame finished the season as a better team than Miami with a better résumé. Miami got in ahead of them. The sole justification was the head-to-head win. Tastes differ on whether this was a good or bad decision. It was the decision, though. We need to put that into our model.
3. The Committee Is More Inconsistent Than We Thought Possible.
For a long time now, we’ve hoped that the committee would establish some sort of precedent and follow it. This year threw that out the window. If the committee can’t use a consistent rubric within one set of rankings, let alone one season, we have to build that into our expectations. The direction of the inconsistency is probably going to be predictable, and I unfortunately don’t think we can model that just yet. But we need to leave a lot fewer probabilities at 100% and a lot fewer probabilities at 0% in our model’s final readouts going forward. There needs to be a randomness variable that goes in at the end. We used to have one. We thought it was misleading, giving teams false hope where there was no hope to be found. It turns out it’s necessary. Mammals are an unpredictable Class.
**
We’ll keep updating our model as the College Football Playoff and bowl season unfold. I don’t know yet what this offseason will look like. We’d always like to share updates as rosters finalize and the season gets closer, but generally, we end up scrambling to get this thing ready before Week Zero. We’ll keep you posted. In the meantime…here comes college basketball. Time to get that NIT Bracketology back up and running.
**
