The College Football Playoff Selection Committee released new rankings last night, which were, of course, swiftly met with outcry. Is the outcry justified? Do these rankings really matter? Here’s what we know, with a lot of help from our model:
A Long, Technical Explanation About This Piece of Our Model
Our model’s formula for predicting the committee’s rankings is based on the five years of final rankings we have available. One variable in that formula is something we call a “Forgiveness/Punishment Adjustment.” We’ve been struggling to figure out a short name for it, so today, let’s call it FPA.
FPA’s purpose, when we designed it, was to account for factors like, say, a playoff contender losing a game in which its starting quarterback left in the first half with a concussion, or a playoff candidate losing so badly on the road that the margin began to matter in addition to the loss itself. We realized once we had the variable that it could also be an effective way to adjust our model to account for each round of the committee’s rankings. We do believe the committee’s rankings are more on the formulaic side than the horse race side, which means we think the committee really is evaluating total body of work each week, rather than approaching their task through a lens in which the previous week’s rankings are the starting point and only the most recent results can change things. Still, though, our formula can only go so far, and FPA allows us to recognize how the committee is evaluating teams’ respective résumés at this point in the season and adjust our model to account for that. It also reflects how much forgiveness, reward, or punishment the committee is granting teams relative to the precedents set by prior committees.
None of this is to say the committee is right or wrong about anything, or that prior committees were right or wrong. FPA is simply a number we can generate that reflects how and where the committee is deviating from trends established in previous seasons. A positive FPA doesn’t mean the committee is over-ranking a team—it just means the team is being ranked more highly than we’d expect given precedent. A negative FPA doesn’t mean the committee is under-ranking a team—it just means the team is being ranked lower than we’d expect given precedent. Of course, the committee may be over and under-ranking teams, but that’s subjective, though I will offer that FPA can be a useful tool in arguing how over or under-ranked a team is.
Our model applies FPA each week once the rankings are released, by going through the rankings from the top and adding or subtracting FPA until the model’s expected rankings at this point in the season line up with those of the committee. Many teams receive an FPA of zero. When a team receives an FPA, it sticks around going forward. A team’s FPA can be changed week-to-week in accordance with the committee’s rankings. FPA can’t perfectly match its own real value (effectively the value of all the unknowns impacting the committee’s collective opinion at this point in time), so it errs on the side of being as small as possible. In each of our model’s simulations, a degree of randomness is added to reflect the possibilities of FPA’s changing and the reality that new results will continue to affect FPA’s.
That’s a lot of wonky stuff. The real message here is this: Our model is measuring the difference between how the committee is ranking teams and how our model expects the committee to rank teams, then folding that into its simulations. More simply, this is how the committee’s rankings this week affect our understanding of the playoff picture:
Starting at the Top
Ohio State led our model’s expected rankings by a large margin entering this week: on a scale in which the top-ranked team sits at 100 and UMass (the bottom-ranked team in the FBS) sits at zero, Ohio State led LSU by about 8.5. Yet, LSU was ranked ahead of them.
What’s happening here seems to be that LSU is being given more credit than past committees have given for their best win, this weekend’s victory at Alabama. Simultaneously, Ohio State is being given less credit than previous committees have given for systematically demolishing all of their opponents relative to those opponents’ average results.
This did drop Ohio State’s playoff chances a touch, from 98.8% before the rankings were released to 94.6% after the rankings were released, but the Buckeyes’ national championship chances are still about 50%, reflecting how the most likely scenario our model foresees is now one in which Ohio State’s chances of making the national championship rise by getting to play Clemson in the semifinal rather than possibly Alabama—this shift from prospective competition evened out the shift from playoff probability, which is worth noting. LSU’s playoff chances, of course, also rose on the news, from 86.2% to 92.8%.
LSU is the only team in the SEC West with better than a 0.3% chance of winning the conference, which means they’re all but certain to make the SEC Championship. The 7.2% of simulations in which they don’t make the playoff almost all involve both a stunning regular season loss and a loss to Georgia in the conference title game. Lose just once, and the Tigers are probably in.
LSU’s FPA this week, scaled on that same zero-to-one-hundred scale that measures the gap between them and UMass, is 5.3. Last week, it was 2.4, so it increased by 2.9 this week.
Ohio State’s FPA is now -3.2, after sitting at 0.0 last week.
The Committee Loves Georgia
Georgia’s FPA is the highest in the country, now sitting at 7.3 after entering the week at 5.1. Our model knew the committee was high on the Dawgs, but it didn’t realize how high was high. Georgia’s now 33.2% likely to make the playoff. 24 hours ago, that occurrence was only 23.0% likely.
Alabama’s Actually Being Punished
A lot of the discourse I personally saw criticized the committee for keeping Alabama in the top five, citing Alabama’s lack of marquee wins. Our model accounts for this, though, and it was still surprised to see the Tide drop as low as they did.
Alabama’s best win, to date, is a road victory over Texas A&M. By our model’s scale, this is better than the best wins of Ohio State (Wisconsin at home), Clemson (ironically, UNC on the road), or Georgia (Florida at a neutral site). Their second-best victory (over South Carolina on the road) is worse than those of all four teams above them, but still, folding everything in, our model indicates former committees would have the Tide ahead of the Dawgs by a wide margin. After coming in last week with a mildly positive FPA of 0.6, theirs now sits at -1.8.
For what might Alabama be receiving this punishment? The most objective possibility is that the committee believes their nonconference schedule is weaker than necessary. A more subjective one is that the committee does not want to be seen as giving Alabama the benefit of the doubt. Regardless of their rationale, the rankings dropped Bama’s playoff probability from 41.8% to 21.0%. That matters, and it helps the national championship chances of every other team.
Utah Got Some More Love
In keeping with their theme of forgiving unsavory losses, the committee continues to seemingly downplay the importance of Utah’s defeat at the hands of a banged-up USC. Their FPA entered the week at 1.9. It rose again to 2.9.
Penn State Got Punished
In keeping with their theme of taking it out on teams who lost to undefeated opponents, Penn State received a large negative FPA this week: -4.4, currently the largest of all negative FPA values. Our model thought they were 16.9% likely to make the playoff entering the rankings. Now, they’re 6.4% likely.
Minnesota’s Sins Were Forgiven
Last week, Minnesota’s -3.6 FPA was a story. Their triumph Penn State erased the whole thing. The Gophers’ FPA is now an even zero, and they’re edging into the hunt. A 4.0% playoff probability isn’t much, but it’s worth getting excited over, and unlike those of its neighbors, their probability is one that can rise based solely on their results. In other words, Minnesota doesn’t need any help.
Nothing New for Baylor, a Windfall for Texas, and a Blind Eye towards OU
While Texas received a 4.1 FPA, the largest doled out this week, Baylor’s sat constant, at -3.6. The Bears’ abundance of narrow victories, paired with a soft non-conference schedule, is our subjective, personal guess at what’s holding them back. As Minnesota showed, though, that can be forgiven in an instant.
What can also evidently be forgiven is a narrow home escape against Iowa State, because while our model expected Oklahoma’s performance Saturday night to bump the Sooners down a few pegs, they didn’t budge, with a newfound 1.5 FPA holding them aloft.
What’s Going On, Overall?
At the highest level, the playoff picture now breaks down as follows (playoff probability in parentheses):
Teams who control their fate and are expected to do so:
Ohio State (94.6%)
LSU (92.8%)
Clemson (86.3%)
Teams who may control their fate, but may not, and aren’t expected to do so:
Georgia (33.2%)
Utah (20.4%)
Oklahoma (17.4%)
Oregon (14.2%)
Penn State (6.4%)
Minnesota (4.0%)
Baylor (1.9%)
Teams who don’t control their fate, but could make it with varying degrees of help:
Alabama (21.0%)
Florida (3.9%)
Wisconsin (1.4%)
Auburn (1.1%)
Michigan (1.0%)
***
More to come on Friday, where we’ll break down all Saturday’s big games and their playoff implications. Come back and see us then.
