We’ll get to our reflections on our model, and on Tuesday’s rankings, and on our model’s big miss on those rankings, but we’re going to do that later, down below, because it isn’t exactly the most important thing right now. The most important thing? Who actually makes the playoff.
The Scenarios
It’s possible we’ll see Georgia or Michigan blown out this weekend, and if we do, a reevaluation will be in order, but it’s extremely unlikely it happens. Losses by those teams are possible. We’ll give a number below on how possible losses by those teams are. But in our model’s latest 10,000 simulations, Georgia only missed the playoff four times, and Michigan only missed on six occasions. Better odds than winning the lottery, but close to the order of magnitude of the probability of getting struck by lightning at some point in your life.
So, then, it’s two open spots, and the candidates are TCU, USC, Ohio State, and mayyyybe Alabama (and mayyyyyyyyyyybe Tennessee, who made the field in five of those 10,000 simulations, but—again—lightning). With the Buckeyes and Tide both idle, that leaves us with two games that impact the College Football Playoff’s final four: The Pac-12 Championship tonight and the Big 12 Championship tomorrow in the midday.
If TCU wins, TCU is in the playoff. If USC wins, USC is almost assuredly in the playoff. Our model sees a few scenarios where it doesn’t happen, scenarios where USC both wins the Pac-12 and gets jumped by Ohio State, but frankly, that’s a shortcoming with our model, and one we need to address this offseason (more on that below). It’s possible it’d be justified—if Michigan beats Purdue by 200 points and USC only beats Utah by one after Utah lets its drum major play quarterback as part of a surprise performance art installation, the committee would be well within its rights to say, “You know, Ohio State’s probably the better team here”—but it isn’t going to happen. The committee is not going to reevaluate the USC/Ohio State question in any scenario where USC wins.
Similarly, if USC loses, USC is out. Or, I’m about to go Howard Beale on the blogosphere. The only scenario where an 11-2 USC could make the field would be one where either they somehow stay ahead of Ohio State (and it’s already stupid that they’re ahead of the Buckeyes) or one where TCU is pulverized so thoroughly that the Frogs fall past the Trojans. In that latter scenario, the conventional wisdom seems to be that Alabama will be ahead of both TCU and USC, but our model would be stunned if the Tide jumped a 12-1 TCU team.
Really, TCU is the team who can lose their game and still make the field. If USC wins and TCU loses, no, the Frogs probably don’t make the cut. It’s probably Ohio State (though that’s not guaranteed). If USC loses, though, TCU moves very close to being a lock. It’s been a lot of close games. They’ve rarely won impressively. Their best wins aren’t sensational. But 12-1 is 12-1, and if the committee thinks this highly of this 11-1 USC (which is basically TCU with a brand and a loss), it would constitute high hypocrisy to punish a 12-1 TCU to that degree.
So, what we’re guessing the scenarios are…
- USC wins, TCU wins: TCU and USC make the playoff
- USC wins, TCU loses: USC and Ohio State make the playoff
- USC loses, TCU wins: TCU and Ohio State make the playoff
- USC loses, TCU loses: Ohio State and TCU make the playoff, but Alabama watches their television quite closely come Sunday
The Games
All ten conference championships, in order of viewing interest. SP+ and FPI are ESPN systems. Movelor is ours. The betting market win probabilities are implied from available moneylines. The SP+ win probability is our estimation based on its spreads.
Pac-12 Championship: USC vs. Utah (Friday, 8:00 PM EST, FOX)
- Movelor: Utah -4.4 (62.2%)
- FPI: Utah -3.9 (61.2%)
- SP+: Utah -0.3 (50.9%)
- Betting Markets: USC -2.5 (57.3%)
How good is USC? We’re still asking it, and it’s still important. The Trojans’ best victories came over Oregon State, UCLA, and Notre Dame, two of those by only a field goal and none by as many as two touchdowns. Their average margin of victory in conference wins was merely 13 points, and that number drops to nine if you exclude the game against Colorado. Bowl-ineligible Cal, Arizona, Arizona State, and Stanford? USC won those games by an average of eleven points.
It’s very much the team sheet of a crew which played to its level of competition, and the problems with that sort of evaluation are 1) it’s unlikely to be true, that is not a sustainable long-term trend; 2) it sort of turns every game into a tossup; and 3) at some point, the team involved will run into someone great, will not play to their level, and will finally answer our question of how good the team actually was.
So, no, we don’t know how good USC is. But we do know that Utah is a good-not-great team. We’ve settled there with them. They’re also inconsistent, but they’re less one-dimensional, so their scores are more predictable (one explanation of USC is that Caleb Williams and that dynamite offense—which really is dynamite—just score however many points they need to while the defense picks dandelions in the outfield). They should be better than USC, but markets don’t agree with that, and to some extent, that speaks to the weakness of college football projection systems. A 6.5-point gap between median rating system projections and betting markets doesn’t happen in college basketball in games of this importance.
In the end, it appears to boil down like this: USC will score. Will Utah score enough? It doesn’t have to come through sensational offense by the guys from Salt Lake. It can come by beating a bad defense. But that’s the question.
Big 12 Championship: TCU vs. Kansas State (Saturday, 12:00 PM EST, ABC)
- Movelor: TCU -3.1 (58.7%)
- FPI: TCU -0.9 (53.9%)
- SP+: TCU -2.9 (58.2%)
- Betting Markets: TCU -1.5 (52.4%)
If your takeaway from these numbers is that betting markets may be undervaluing the likelihood of TCU winning by exactly one point, join me, because I’m seeing the same thing. But I digress.
Everyone agrees on this one: We don’t know what’s going to happen, but TCU’s better by at least a little.
A question that never really gets answered with college football playoff selection is how much better (or worse) teams can and do get over the course of a season, and how much that should be allowed to matter. It’s not the most relevant with TCU, but it does have something to say about this game. Because both TCU and Kansas State seem to have gotten a lot better as this year’s gone on.
Maybe this is unfair to K-State’s schedule. Their early loss to Tulane looks fair in the eyes of the committee, and their losses to TCU and Texas were likewise fair play. But sometime around that TCU game in mid-October, the K-State offense found another gear, and it was at a similar point in the year that TCU’s defense started to figure things out. Texas’s offense isn’t sensational, but TCU holding it to a field goal changed things, and Iowa State had a bad year, but TCU beating the Cyclones over the head with their own bones was a convincing display of force. These teams, Kansas State and TCU, were solid when they played in October. I do think they’re better now.
The question that comes from that question is, Which is better-er? TCU’s spent the last three weeks alternatively impressing and playing with magic beans. K-State lost to Texas a month ago but answered that by punking Baylor but answered that by letting their foot off the defensive gas more than you’d like to see.
It’s a toss-up’s toss-up. Maybe that TCU-by-1 idea really is the call.
SEC Championship: Georgia vs. LSU (Saturday, 4:00 PM EST, CBS)
- Movelor: Georgia -14.6 (83.9%)
- FPI: Georgia -12.6 (83.0%)
- SP+: Georgia -19.1 (89.7%)
- Betting Markets: Georgia -17.5 (88.0%)
This game’s in Atlanta, and I don’t know if that gives Georgia an advantage or not, but none of the rating systems appear to be incorporating that into their math (I know ours isn’t).
This game matters not in determining who makes the playoff (Georgia, as we said above, is almost assuredly in), but in giving us a read on the Dawgs ahead of their national championship push. Movelor took the step this week of pushing Michigan ahead of UGA in rating, and while it’s alone in that among the three systems we reference, it does go to show that Georgia has been off lately. They struggled to put away Kentucky. They struggled to pull away from Georgia Tech. And as we unwind and zoom out and look at their schedule, there’s a rather compelling case that there weren’t great teams this year in the SEC or the Pac-12. Tennessee’s win over Alabama? Narrow and at home, for one thing, but also over an underwhelming Alabama. Oregon, whom Georgia obliterated in this stadium in September? The what, fourth-best Pac-12 team?
The case that Michigan is the best team in the country isn’t one that rests on Michigan having caught up to Georgia. It’s one that rests on Georgia having receded towards Michigan, but having receded more slowly than other tides (including one Tide). Against an LSU team capable of holding opponents down in the muck now and then, we want to see whether Georgia’s again-questionable offense can pull out some fireworks and send a message. If they don’t, we could be looking at the most interesting national semifinal round in five years.
Big Ten Championship: Michigan vs. Purdue (Saturday, 8:00 PM EST, FOX)
- Movelor: Michigan -21.7 (92.1%)
- FPI: Michigan -17.4 (89.0%)
- SP+: Michigan -24.3 (94.0%)
- Betting Markets: Michigan -16.5 (87.7%)
This is the companion piece to the Georgia game, and in scenarios where USC and TCU have each won, we may know by this point that Georgia and Michigan are on a direct collision course.
Blake Corum’s injury is a big deal for college football. It’s a big deal. Heisman implications! But for how good Michigan is…did you see them on Saturday? It’s only one data point, but the implication we’re getting is that Michigan has the backs to fill in adequately for Corum, and that the real strength of this team (and of Corum) lies in the offensive line play anyway, which is something we already knew.
The other conspicuous data point involving Michigan comes from two weeks ago, when they nearly lost to Illinois, and I think we’re still trying to figure out what that was. Was Jim Harbaugh sandbagging, keeping things in the pocket for Ohio State? Did Illinois play its game of the year and then some, known good-coach Bret Bielema making enough progress to compete with the biggest boys in town? Is Ohio State quite bad and is the committee right about everything and should I start writing Boo Corrigan letters asking for advice in all areas of my life?
We might get some answers here, not because Purdue’s great or anything but because Purdue is notoriously feisty. Purdue has a reputation for causing trouble, and that reputation will probably not come through here, but it isn’t that hard to picture Aidan O’Connell pulling a five-touchdown game out of his knapsack and Michigan needing to play its very best bully ball to set up a kick to win 38-35.
We also might get no answers, because Purdue is a fairly mediocre team who had the good fortune of playing neither Michigan nor Ohio State in the regular season and of playing Penn State at home in each team’s opener.
We might be shown that Michigan’s the real deal. We might be shown that Michigan’s a good team who had a good day, but isn’t much more. The trick is, Purdue’s inconsistency means we won’t truly know which it was. Possibly ever.
AAC Championship: UCF @ Tulane (Saturday, 4:00 PM EST, ABC)
- Movelor: Tulane -6.6 (68.0%)
- FPI: Tulane -2.9 (59.3%)
- SP+: Tulane -4.6 (62.8%)
- Betting Markets: Tulane -4 (62.7%)
When we say this holds lower viewing interest than Georgia/LSU, we say that mostly out of obligation. This should be the third-best game of the weekend, combining stakes with drama. UCF and Tulane aren’t good, but they aren’t bad either, and while there was some bullshit which led to this becoming the Cotton Bowl play-in game for the Group of Five (UCF is overranked), it’s fine to give that spot to whoever wins this. There isn’t an obvious alternative team getting snubbed. UTSA has a lot of close wins against teams in Conference USA.
Tulane is very much a “No, but seriously” team, by which I mean that every time they won a surprising game this year, many of us said, “No, but seriously,” and then waited for a fall on their faces that never really came. We thought it had come when they lost to Southern Miss immediately after beating K-State, but then they went and beat Houston, and then they went and beat ECU, and even after they lost to UCF in these teams’ first meeting they turned around and stomped SMU before upsetting Cincinnati.
UCF isn’t as surprising, but they’ve been a team who won their big games and lost some others, with the one exception being that they let Louisville win in Orlando back in their season’s second game. Cincinnati? SMU? Tulane? Beat them all. ECU? Navy? Lost, and in the ECU game lost in a blowout.
The fact UCF’s beaten Tulane once already probably says something, but it’s hard to know what, and whatever it says, markets and systems seem to be in agreement that it isn’t enough to make UCF the favorite.
ACC Championship: Clemson vs. North Carolina (Saturday, 8:00 PM EST, ABC)
- Movelor: Clemson -10.9 (77.5%)
- FPI: Clemson -10.4 (77.7%)
- SP+: Clemson -9.9 (75.5%)
- Betting Markets: Clemson -7.5 (72.4%)
The next-best mid-major championship is the ACC’s, where Clemson is trying to hold onto one last shred of its era as a national power and UNC is trying to at least regionally arrive at long last. This is a good time to point out that reliably dominating one’s conference is a valuable skill in the college playoff world, and also that in the twelve-team playoff, Clemson could get a first-round bye with a win in this game, which is silly but also fine, as they would then be trounced in the quarterfinal by Ohio State.
Mountain West Championship: Fresno State @ Boise State (Saturday, 4:00 PM EST, FOX)
- Movelor: Boise State -6.6 (67.9%)
- FPI: Boise State -6.7 (69.2%)
- SP+: Boise State -5.0 (63.8%)
- Betting Markets: Boise State -3 (59.2%)
If you follow this stuff closely, you may remember that folks were writing eulogies for the Boise State dynasty back in September, when the Broncos were smacked by UTEP and fell to 2-2. Since then, they’ve won seven of eight, losing only to BYU and only by a field goal in that game. They aren’t the nation’s best mid-major (that’s probably Tulane), but they’re still in the mix, and with Idaho’s population growing and the Pac-12 trying to expand, the potential remains there for long-term national relevance.
Which would make it pretty fun for Fresno State to knock them off, I’d imagine.
Sun Belt Championship: Coastal Carolina @ Troy (Saturday, 3:30 PM EST, ESPN)
- Movelor: Troy -8.0 (71.3%)
- FPI: Troy -8.7 (73.9%)
- SP+: Troy -10.0 (75.7%)
- Betting Markets: Troy -8 (73.7%)
There’s some sad winking and nodding going on here, because James Madison’s probably the best team in the Sun Belt but is in the midst of their transition to the FBS level, so there’s no chance to prove it yet. Troy—and the Sun Belt as a whole, really—is in a weird spot where because they occupy the same region as the SEC, they should theoretically have access to much of the same talent. The potential is teeming, but so very hard to really touch. A win by the Trojans would give them a shot at the first twelve-win season in school history. A win by the Chanticleers would give them a shot at their third straight eleven-win effort.
MAC Championship: Toledo vs. Ohio (Saturday, 12:00 PM EST, ESPN)
- Movelor: Ohio -6.6 (67.9%)
- FPI: Toledo -1.9 (54.5%)
- SP+: Toledo -0.9 (52.5%)
- Betting Markets: Toledo -3 (57.1%)
Ok, honesty hour, I’m putting this ahead of Conference USA because last year they cut straight from College Gameday in Atlanta to a clip of the MAC trophy on an ice rink in Detroit and a friend of mine still texts me the picture here and there (thanks, Derek, you’re the best). I’m ashamed to admit that I don’t remember who won the last MAC Championship, or any MAC Championship since Western Michigan in…2016? I remember them and I remember Northern Illinois in 2012 and I know there was a fun Ball State but I’m not positive they won the MAC. Did Buffalo win it last year? It was probably Buffalo, right?
Anyway, apologies to the MAC, you deserve better, but Ohio’s got a fun offense and Toledo’s been in that nationally respectable window in both football and men’s basketball at points in the last decade, and we’re down with all that. Give us some MACtion. (Also, what’s up with Movelor having three lines at exactly 6.6? We good?)
Conference USA Championship: North Texas @ UTSA (Friday, 7:30 PM EST, CBSSN)
- Movelor: UTSA -12.6 (80.7%)
- FPI: UTSA -9.0 (74.3%)
- SP+: UTSA -11.2 (78.1%)
- Betting Markets: UTSA -8.5 (74.0%)
UTSA opened the season with a dramatic, three-overtime, two-point loss at Houston. Two weeks later, they fell at Texas by three touchdowns in a game the Roadrunners led by ten midway through the second quarter. Since that defeat in Austin, it’s been all UTSA, and while numerous foes played them close—including UNT, against whom UTSA had to score twice in the final three minutes to escape—they won them all. One more now, for a Conference USA sendoff before they jump to the AAC this summer.
Rankings Talk, Model Talk
Ok, those were the games. You want to hear about the rankings?
I accidentally preemptively wrote this on Tuesday, but I cannot fathom the committee ranking this 11-1 USC ahead of this 11-1 Ohio State. There’s a case to be made for the hypothetical 12-1 USC, especially in a world where the playoff is optimally designed and teams who’ve lost decisively to other playoff teams are excluded, but Ohio State is better than USC in every category except margin of defeat and recency of defeat, the latter of which shouldn’t be a category.
But, it’s not our model’s job to predict rankings I can fathom. Our model’s purpose is to predict the committee’s rankings, and at that, it loudly failed. We were wrong about USC and Ohio State.
There’s a clear culprit here, and there’s also a suspected accomplice. The culprit is how we go about FPA, our model’s gauge of how much the committee’s rankings deviate from precedent. Our FPA method is clunky and overly simplistic, and that’s always been fine but it needs to get better.
We still don’t believe the rankings process is a horse race, the way the AP Poll was. The horse race elements, though, are louder than we’ve conceded. We need to adjust FPA’s randomness in our model’s simulations to account for this. They need to correspond with various levels of surprise in results, and they need to correspond with recency, and in areas where we don’t have precedent, they need to err on the side of being uncertain. This is a big offseason project.
The suspected accomplice is APD, Adjusted Point Differential, our method of measuring how convincingly teams win. By APD, Ohio State has won more convincingly than any team in the country, measuring by simply subtracting each game’s margin from the opponents’ season-long average margin (so if Ohio State beat Iowa by 40 and Iowa averaged winning by 10, Ohio State would get credit for a 50-point overperformance in that game). APD plays a big role in our model, and it’s never been a problem before, but it’s what created the gap between Ohio State and USC, because say what you will about Caleb Williams and the Heisman but USC has not won convincingly.
This last bit gets us into more unsavory territory, which is the subjectivity of the whole enterprise and the importance of branding. In college basketball, Michigan State memorably made the 2021 NCAA Tournament seemingly because Tom Izzo is Tom Izzo and Michigan State is Michigan State. Something similar appears to be happening with USC in these latest rankings. All season, we’ve heard TCU questioned for not winning convincingly, but now that USC’s reached eleven wins, winning convincingly doesn’t matter? Is it because they’re USC? Is it because Lincoln Riley is Lincoln Riley? Is it because Caleb Williams is in the Heisman picture? This is the part that sucks, and the thing that really pisses us off about it is that we might have to revisit the question of whether the AP Poll is a useful variable for our model to include.
FPA numbers, from the rankings, positive means the team was forgiven for poor performance, relative to precedent, negative means the team was punished, relative to precedent:
- UCF: +9.4
- Kansas State: +3.6
- NC State: +3.4
- South Carolina: +3.0
- UCLA: +2.6
- Tulane: +2.6
- UNC: +1.9
- TCU: +1.7
- USC: +1.7
- Florida State: +1.7
- LSU: +1.7
- Washington: +1.1
- Utah: +0.9
- Georgia: 0.0
- Oregon State: -0.2
- Clemson: -0.4
- Mississippi: -0.6
- Purdue: -0.6
- Wake Forest: -0.6
- UTSA: -0.6
- South Alabama: -0.9
- Michigan: -1.1
- Troy: -1.5
- Oregon: -1.9
- Notre Dame: -2.3
- Alabama: -2.8
- Mississippi State: -2.8
- Penn State: -3.0
- Ohio State: -3.8
- Tennessee: -3.8
- Texas: -9.0
Yes, Ohio State and Tennessee are, to our model, two of the three most surprisingly punished teams in the country. Yes, the committee appears to be lower on the SEC than precedent would suggest. Yes, the ACC continues to be treated like an adult. Yes, UCF’s ranking is still batty.
Two last thoughts, before we move on:
First, Georgia did not receive this treatment when Alabama beat them up in the SEC title game last season. Some fairness exists there (the SEC is better than the Big Ten), and some of it is convenience (there was no strong alternative at that point), but treating the Big Ten as equivalent to the Pac-12 is stupid. It’s just stupid.
Second, maybe the committee wanted to signal that USC’s in with a win tonight. Maybe they were trying to eliminate any debate. That might be the best I can come up with. It’s wacky, and it’s wild, but it’s temporary enough (thanks to the aspect where USC will either win its conference championship and move half a game ahead of Ohio State in win-loss record or lose and become irrelevant to the conversation) that it’ll probably get passed over when we look back on playoff fields. Our task will be making sure our model doesn’t forget.
