SportsRatings: College Football

Playoff algorithm? Who knows, it's all up to the humans now

Florida State and Michigan won, but after the Wolverines and Huskies it's a crapshoot due to unpredictable human beings.

2023 Playoff Odds 12pm; after all games played but before AP poll

 Ranking                                   | Playoff    | CFP   
 Prv Rnk Team                   Conf   Rec |%Odds  %Chg | LW  
  2   1. Michigan               B10e   13-0  99.9  +2.5    2   
  1   2. Washington             P12    13-0  98.7  +0.3    3   
  5   3. Florida St             ACC    13-0  78.0  +5.7    4   
  4   4. Texas                  B12    12-1  48.7 -15.8    7   
  6   5. Alabama                SECw   12-1  39.3 -14.4    8   
  3   6. Georgia                SECe   12-1  28.0 +15.3    1   
  7   7. Ohio State             B10e   11-1   7.4  +6.4    6   
  8   8. Oregon                 P12    11-2   0.0   0.0    5   

Well here we are, our formula reduced to logic and precedent rather than a combination of odds and outcomes. Logic dictates that Texas is ahead of Alabama who is ahead of Georgia. Will that logic hold with the humans in the Committee? If some of the talking heads are correct, then it might not. Precedent states that a 13-0 power conference champ is guaranteed a spot. Will that precedent hold? Again, the talking heads are saying, maybe not.

Will Florida State make it? Will Texas make it? Or will the idea that the SEC is the best conference—not true this year, btw—mean that both Alabama and Georgia make it? Our projection can't deal with this kind of sentiment, and simply projects the three 13-0 teams to be at the top, joined by the team that beat the team that beat the team. Texas isn't that close to jumping FSU, and Alabama isn't that close to jumping Texas. In fact it's surprising how close Georgia is to being ahead of Alabama. And Ohio State still gets consideration for being 11-1, maybe being the best team in the country, and only losing on the road to the #1 team. That's as good as Georgia's résumé. 

But it's not about résumé at this point. This year is looking a lot like 2014, when there were six teams to fill four slots. That year an undefeated Florida State made it in at 13-0, instead TCU was sacrificed to open a spot for Ohio State. Nobody had a 59-0 win this year, but people are acting like Alabama's 24-21 win over Georgia is the same thing. 

Florida State's standing doesn't capture the fact that humans will look at: a team without their star quarterback. They very well might decide the Seminoles are not "one of the four best teams" and also decide the SEC champ can't be left out. Having decided that, they better put Texas in. Including the team that beat Alabama instead of the team that lost to Alabama would at least be a nod toward a semblance of fairness.

I'll revise these projections after the AP poll comes out, as the sportswriters are on the same page as the Committee and what they decide is probably quite close to what the Committee will decide.

Georgia loses, FSU looks in if they win; Texas 4th, Alabama 5th

Alabama beat Georgia and the team helped most in our sim was Florida State, who now looks in if they win.

2023 Playoff Odds 7:30pm 12/02: after Pac-12, Big Twelve, SEC championships; pre-Big Ten, ACC

 Ranking                              Curr | Playoff    | CFP   Chalk
 Prv Rnk Team                   Conf   Rec |%Odds  %Chg | Curr  Proj
  1   1. Washington             P12    13-0  98.4   0.0     3     2  
  2   2. Michigan               B10e   12-0  97.4  +1.0     2     1  
  5   3. Florida St             ACC    12-0  72.3 +28.6     4     3  
  4   4. Texas                  B12    12-1  64.5  +7.8     7     4  
  6   5. Alabama                SECw   12-1  53.7 +29.0     8     5  
  3   6. Georgia                SECe   12-1  12.7 -58.9     1     6  
  7   7. Ohio State             B10e   11-1   1.0  -7.0     6     7  
  8   8. Oregon                 P12    11-2   0.0   0.0     5     8  

Why was FSU the big recipient of Alabama's win? In our sim at least, clearing out undefeated teams let FSU gain a higher ranking if they win, keeping them away from the dreaded #4 spot which Texas has to fight out with Alabama. If FSU loses, they should both get in comfortably ahead of Georgia, otherwise the Committee will have its say.

The SEC can still get either shut out, or get two teams in if FSU and Michigan both lose. That would be a chaotic situation but Alabama and Georgia might finish #3 and #4 then.

Texas wins 49-21, Playoff odds dip?

Texas walloped Oklahoma State 49-21, a 35-point win that wasn't enough to clear the bar for our algorithm to pass Florida State if the Seminoles win by more than 4 points.

That may seem arbitrary and it kind of is, and other factors change how our model reacts too. For one, the expected Committee ranking becomes more and more important, and that helps Florida State a bit and hurts Texas.

2023 Playoff Odds 3:30pm 12/02: after Pac-12, Big Twelve championships

 Ranking                              Curr | Playoff    | CFP   Chalk
 Prv Rnk Team                   Conf   Rec |%Odds  %Chg | Curr  Proj
  1   1. Washington             P12    13-0  98.4   0.0     3     3  
  2   2. Michigan               B10e   12-0  96.4  +0.7     2     2  
  3   3. Georgia                SECe   12-0  71.6  +0.6     1     1  
  4   4. Texas                  B12    11-1  56.7  -7.5     7     5  
  5   5. Florida St             ACC    12-0  43.7  +5.7     4     4  
  6   6. Alabama                SECw   11-1  24.7  +5.9     8     8  
  7   7. Ohio State             B10e   11-1   8.4  -5.4     6     6  
  8   8. Oregon                 P12    11-2   0.0   0.0     5     7  

Michigan and Georgia stayed about the same, as neither was threatened by Texas. The Longhorns taking a 2nd loss would have seen a bigger improvement since a 1-loss Texas hurts the Wolverines and Bulldogs chances at 12-1. So this rise is mostly from slight formula changes as the Committee projection itself grows more important than other performance factors (also since it's a Monte Carlo simulation, this could be a more or less random result).

Texas should have gained and FSU lost odds, you would think, but the opposite happened. As we covered before, the amount Florida State wins by affects the result, but the balance of factors changes slightly with more information. As it is, Florida State should get in—in our simulation—if they win by a touchdown or more. In real life, any win should be enough but the big Texas win might make the Committee consider FSU's QB situation and whether they're one of the four best teams if they win narrowly.

Alabama is a theoretical threat to both Texas and Florida State, and their odds get a boost, too. But that's mainly taken from Ohio State's diminishing odds. With Texas at 12-1, that's another team guaranteed to be above the Buckeyes. There's no path for them to pass Texas anymore, and with Washington ahead of them, they'd have to pass a 12-1 Michigan (unlikely) or stay ahead of a 12-1 Alabama (also unlikely). Their odds here essentially are for the many simulation runs where they finish #5, which gets credit for being close due to Committee uncertaintly. But when push comes to shove the Buckeyes are almost certainly out of luck, and a 12-1 Alabama captures most of their lost odds. But they still don't get in with a win right now on most trials.

After Washington win, new Playoff Odds help Texas, Alabama

Washington beat Oregon for a 2nd time and their odds predictably took a big jump, making them currently the most likely team to make the Playoff.

Seems like a no-brainer obviously. The Huskies don't have 100% odds only because they normally finish 3rd in the simulation and there's still some uncertainty built in for any team that doesn't finish at the top right now.

Michigan's odds are almost as high as Washington's even though they have yet to play Iowa. Their number barely budged from the Pac-12 game result as they are still ahead of the Huskies if they win, which is about a 95% chance according to our numbers.

2023 Playoff Odds 11:00am 12/02: post-Pac-12 Championship game only

 Ranking                              Curr | Playoff    | CFP   Chalk
 Prv Rnk Team                   Conf   Rec |%Odds  %Chg | Curr  Proj
  6   1. Washington             P12    13-0  98.4 +64.1     3     3  
  1   2. Michigan               B10e   12-0  95.7  -0.1     2     2  
  2   3. Georgia                SECe   12-0  71.0 -11.1     1     1  
  5   4. Texas                  B12    11-1  64.2 +12.5     7     5  
  4   5. Florida St             ACC    12-0  38.0 -16.3     4     4  
  8   6. Alabama                SECw   11-1  18.9 +11.2     8     8  
  7   7. Ohio State             B10e   11-1  13.8  +4.7     6     6  
  3   8. Oregon                 P12    11-2   0.0 -65.1     5     7  

Georgia took a small hit from the result, as they might finish below a 13-0 Washington but were always ahead of a 12-1 Oregon if they beat Alabama. In the big scheme it's pretty inconsequential because they're guaranteed a spot at 13-0 anyway, and the 12-1 result depends on what other teams (and the Committee) do.

Alabama's odds improved as Oregon's loss puts them clearly past the Ducks, who decline to 0.0% with their 2nd loss. It also puts their odds higher than Ohio State's odds for the first time. The Buckeyes clear Oregon too but can't do anything to improve their position, and therefore are dependent on many teams losing. Alabama on the other hand either helps or hurts their cause a lot. Oregon's loss puts them more within striking distance if they beat Georgia. But getting past Texas is still a problem.

Speaking of the Longhorns, the biggest thing that happened is that Texas's odds jumped quite a bit while Florida State's declined. For FSU, it added a team that always finishes ahead of them; in scenarios where Oregon beat Washington the Seminoles often finished ahead of both, but they now finish behind Washington in every sim. And since Texas beats Oklahoma State more often than FSU beats Louisville, the Longhorns often take the 4th (or better spot). 

The oddest thing that happens in our sim is that Texas at 12-1 often finishes ahead of 13-0 FSU. That shouldn't happen in our sim, and it shouldn't happen in real life, but there's concern that the Committee might downgrade FSU if they play poorly in a win over Louisville since their starting QB is out for the season. I swear our algorithm is not an AI, and doesn't know anything about the QB position but FSU's lower Strength rating dings the Seminoles sometimes. When FSU only wins by a small margin, 12-1 Texas beats them out. I ran some scenarios and this is what happens:

FSU wins by:    Texas must win by:
      1              1
      2              1
      3              1
      4             14
      5             44
   6-15             44
     16             55   
     17             n/a

Now, these numbers don't have any real-life meaning; they're just edge cases of an algorithm. But they are consistent with human nature in some ways. Remember how Ohio State beat Wisconsin 59-0 in 2014, and suddenly they were in the Playoff? Margin of victory matters.

The numbers here basically say that if FSU wins by a field goal or less, then Texas just has to beat Oklahoma State and they get in at 12-1 ahead of 13-0 Florida State. That seems a bit odd, as Texas winning by a small amount wouldn't make them look very good, either.

A big change happens at an FSU win by 4 points; then, Texas has to win by two touchdowns. I'd say this is true for the 1-3 point FSU win, too.

At 5 points, the FSU win becomes much harder to overcome, and now Texas has to win by 44 points. If FSU wins convincingly, by 16 points, Texas needs a 55-0 win to convince the Committee to go with the Longhorns. And if FSU wins by 17, Texas can't reasonably get the nod.

In our "chalk" projection, FSU gets in because the Seminoles win by around 9 points, which requires a huge Texas blowout and we have the Longhorns winning by about 18 points. So either Texas needs a bigger win, or FSU has to win by 4 points or fewer.

In real life, we assume that a 13-0 FSU gets in every time. And if they don't, it might be a 12-1 Alabama instead of 12-1 Texas, but those cases are very rare in our sim. We can't really simulate the minds of the Committee members; that's why every result has uncertainty build into it, and give the results in what the algorithm determines the odds to be. In FSU's case, their odds of winning stayed the same, but their odds of making the Playoff fell, and it might just happen to coincidentally reflect uncertainty in the minds of the Committee members. We'll see.

2023 Conference Championship game prediction: two upsets (but not the big ones)

Here's what the Strength power rating says about this weekend's final games:

Friday Night

C-USA (7pm Eastern) at Liberty (-10 1/2; 57.5)

New Mexico St. 21
Liberty        34

Pac-12 (8pm) Las Vegas, NV (Oregon -9 1/2; 65.5)

Oregon         37
Washington     26

Friday's games are both fun, featuring two teams trying to finish a perfect 13-0: Washington and Liberty. Washington is shooting for the Playoff while Liberty is gunning for a New Year's Six bowl berth. But Washington is a heavy underdog and we pick Oregon to win the rematch handily, even possibly in a revenge-fueled blowout. Washington will need its close-game magic to win this one. We take Liberty to beat the spread against an amazing 10-3 New Mexico State team (2-10 two years ago), but our power rating routinely exceeds the spread by a few points so these results are both pretty standard.

Saturday

Big Twelve (12noon) in Austin (Texas -15; 54.5)

Oklahoma St.   17
Texas          37

MAC (12noon) at Toledo (-8; 43.5)

Miami OH       19
Toledo         27

One game is important, the other isn't, so most people won't watch the upstart Miami Redhawks (10-2) take on the 11-1 Toledo Rockets, the latter another team gunning for a New Year's Six berth. There we have Toledo winning and matching the spread, but their chance of being the top Group of Five team is minuscule due to Liberty and the ACC title game matchup.

The Texas game is a big deal and becomes a bigger deal if any of the undefeated teams is upset. And really a big deal if Alabama beats Georgia. We give Texas partial home field for playing in state but even without it we'd have them beating the spread (which keeps increasing, up from 12 points to 15). Texas has been pretty consistent, though they keep letting teams back into it after a fast start lately. Oklahoma State is all over the place and if the bad version shows up, 20 points is the minimum loss. If the good Pokes show up, this one will go down to the wire.

Mountain West (3pm) at UNLV (+2; 58.5)

Boise State    31
UNLV           32

SEC (4pm) in Atlanta (Georgia -5; 54.5)

Alabama        24
Georgia        28

The mid-afternoon slate is dominated by Alabama vs. Georgia. The Bulldogs are a 5 point favorite but we have them only by 4 even after giving partial home field for playing in Atlanta. This should be a really close game not because these are the best teams but because both teams are relatively weak this year. Whoever exploits the other team's shortcomings will win. As to what happens if Alabama wins, we have no idea but logic dictates that neither team gets in unless there are other upsets. Whether that logic holds is another matter.

The other game features one of our two upset picks: We have UNLV at home edging Boise State, and here the location makes all the difference in our pick. You have an hour to watch this one to see which way it's going before switching to the really important game.

American Athletic (4pm) at Tulane (-3; 47.5)

SMU            28
Tulane         18

Sun Belt (4pm) at Troy (-5 1/2; 52.5)

Appalachian St 22
Troy           29

The other two games on at the same time as Georgia-Alabama feature our other upset pick, and this one our power rating is pretty emphatic about: we've been picking SMU to win the AAC for half the season now, and still have them winning by 10. The oddsmakers vehemently disagree, making Tulane a 3 point favorite at home. Tulane just hasn't looked very good lately, surviving very close calls (3 points or less) vs. 2-10 East Carolina and 4-8 Tulsa, while SMU beat those teams by 21, and 59. Tulane's only loss was at Ole Miss, while SMU fell at Oklahoma and TCU—all the losses by both teams were by 17 points. The problem for SMU—and the likely reason Tulane is favored, in addition to home field—is that the Mustangs will be without starting QB Preston Stone for the game and his backups have played very little.

Either way, the winner of this game should be the Group of Five pick for the New Year's Six bowl game. If Tulane wins that's a near-certainty; if SMU wins the Mustangs might not get the nod over Liberty since they aren't ranked by the Committee yet, so we don't know where they stand. But it makes sense that they would replace the Green Wave since the Committee likes the AAC and SMU's resume is solid. In the Sun Belt title game we have Troy at home winning as expected, exceeding the spread slightly as our power rating tends to do. The Trojans could make the Committee top 25 with a win but their odds to be the GoF rep are almost zero due to competition from Liberty, Toledo, and the AAC teams.

Big Ten (8pm) in Indianapolis, IN (Michigan -22; 34.5)

Michigan       27
Iowa            3

ACC (8pm) in Charlotte, NC (FSU -1 1/2; 47.5)

Louisville     23
Florida St     31

The last two games, both very important and both with big favorites—at least by our power rating. The spread on the FSU game is very slim, and undoubtedly due to FSU's QB being hurt. Our power rating reflects roughly what the spread would be with Jordan Travis healthy, so he's worth around 6 points to the oddsmakers. But there are rumors about Tate Rodemaker being hurt too, or having suffered a concussion against Florida. And there's the idea that the Committee might pass over a 13-0 FSU if they have a weak win against Louisville, in favor of Alabama if the Tide upset Georgia. And of course if the Seminoles lose, they're out, but Louisville is pretty lousy most of the time, let's face it.

As for Iowa vs. Michigan, we saw the blueprint when Iowa travelled to Penn State and lost 31-0. A 27-3 outcome would be about the same on a neutral field. Michigan's defense is about equal to Penn State's, maybe better, and Iowa doesn't throw anything unexpected that might shake things up. The only question is how much Iowa's defense—which is also as good as Penn State's or Michigan's, roughly—might hold the Wolverines out of the end zone and keep things one late play away from an upset. Odds are the Wolverines grind away and score here and there and become the 2nd team to tally more than 16 against the Hawkeyes.

The only other question is, what are Michigan's chances if they lose? They should remain ahead of Ohio State so they'll be in good shape if either Texas or FSU loses. What hurts a 12-1 Michigan is an Alabama win (which could put two SEC teams in) and an Oregon win (Washington might be ahead of them). But there are lots of paths for Michigan, and by the time of this game we'll know most of the outcomes.