Over the past few weeks we have covered a lot of ground in this series. We started back in 2023, took a close look at the schedule for the Michigan State Spartans, and then made some predictions about the winners of all 10 Football Bowl Subdivision conferences.
Now, it is time to put all of the pieces together and make some predictions about the newly expanded 12-team college football playoff.
Review of Conference Champion Predictions
Table 1 below summarizes the results shared in the previous parts of this series regarding the two teams from each conference who are predicted to face each other in their respective conference championship games.
 |
| Table 1: Review of the predicted conference championship game participants based the preseason simulation results and the disruptive scenario. |
The left side of the table presents the predicted conference championship participants based on the raw odds generated from my preseason Monte Carlo simulation of the full season. The middle of the table gives the results from my "disruptive scenario." This scenario starts by assuming that the favored teams win every single game on the full season schedule, but the results are then adjusted to add a historically reasonable number of road upsets.
The right side of the table gives the record of each team in the disruptive scenario as well as some of the notable wins and losses for each team in that scenario. I generally use the results of the disruptive scenario as my official picks.
Table 2 below shows the teams with the best odds as well as my disruptive picks for each conference champion.
 | |
|
Let's Talk About the Playoffs
If we assume that the season plays out exactly as the disruptive scenario predicts, the four highest ranked conference champions would most likely be:
No. 1 Oregon (12-1 with only a loss at Michigan)
No. 2 Utah (12-1 with a loss at Oklahoma State)
No. 3 Texas (11-2 with losses at Michigan and at Texas A&M)
No. 4 Miami (11-2 with losses at Louisville and Florida)
One could also make the argument for Texas over Utah for the No. 2 seed, but in the new format, this does not make a big difference as we will see later.
The remaining automatic bid in this scenario would be a coin toss between Boise State (12-1) and Appalachian State (12-1) as the next highest ranked conference champion. I will go with Boise State here who would slot into the bracket as the No. 12 seed.
The remaining seven playoff spots would likely be filled with a collection of the following teams with 11 regular season wins:
- Michigan (11-1, with a loss at Ohio State, but wins over Texas and Oregon)
- Mississippi (11-2, with a loss at LSU and to to Texas in the SEC Championship, but with wins over Georgia and Oklahoma)
- Penn State (11-2, with a loss at USC and to Oregon in the Big Ten Championship, but with a win over Ohio State)
- Virginia Tech (11-1, with a loss at Miami, but a win over Clemson)
- SMU (11-1, with a loss at Louisville, but a win over Florida State)
This then leaves two spots available that would most likely go to a Power Four team with a 10-2 record. In this scenario, there are eight teams that fit these criteria:
- LSU (10-2, with losses at Texas A&M and Florida, but wins over Alabama, Mississippi, and Oklahoma)
- Oklahoma State (10-2, with losses at Kansas State and TCU, but a win over Utah)
- Iowa (10-2, with losses to Ohio State and UCLA, and a win over Iowa State)
- Tennessee (10-2, with losses at Georgia and Oklahoma and wins over Alabama and Florida)
- Kansas State (10-2, with losses to Iowa State and West Virginia and wins over Oklahoma State, Kansas, and Arizona)
- Ohio State (10-2, with losses to Oregon and Penn State, but a win over Michigan and Iowa)
- Missouri (10-2, with losses at Texas A&M and Alabama and a win over Oklahoma)
- Kansas (10-2, with losses to Kansas State and West Virginia and wins over Iowa State and TCU)
- Notre Dame (10-2, with losses at Texas A&M and at USC, and wins over Florida State and Louisville.
In order to construct an algorithm that I could use to automatically select at-large teams, I constructed a strength of record metric that compares the number of wins each team earned to the number of expected wins a reference team would have playing the same schedule.
When I used this metric, the teams are listed above in the order which my algorithm would have selected them, with two-loss LSU placing ahead of one-loss SMU and with Oklahoma State claiming the final spot. It seems odd (and unlikely) that Ohio State would be ranked below the Cowboys, but for the this exercise, we will assume the ranking above holds.
With this assumption, the college football bracket is shown below in Figure 1.
 |
| Figure 1: Dr. Green and White's projected 2024 College Football Playoff bracket. |
I have also forecast the results of the tournament using the preseason power rankings and by assuming the projected favorite wins each game. As we can see, I project a few seed upsets in the quarterfinal round (No. 5 Mississippi over No. 4 Miami and No. 7 Michigan over No. 2 Utah) and I have No. 1 Oregon beating No. 3 Texas for the National Title.
Other Playoff Observations
There are several things notable about this bracket which teach us a few things about the new playoff format and about what to expect the inaugural 12-team playoff.
First is the list of teams that are not included in this bracket: preseason No. 1 Georgia, No. 2 Ohio State, and No. 5 Alabama all failed to win their respective conference titles and failed to make the top seven in the final projected rankings in the disruptive scenario.
But this was not because any of those team were worse than expected in my analysis. On the contrary, Georgia and Ohio State would both be favored over all 12 of the playoff participants on a neutral field.
Instead, to borrow a concept from earlier in this series, Georgia, Ohio State, and Alabama were all simply "unlucky" in the disruptive scenario. Those top five teams lost a combined total of eight near toss-up road games.
If Ohio State were to have been included as the No. 11 seed in the bracket above as the "last team in," my analysis would have picked the Buckeyes to have won the National Championship. Instead, this scenario would have Ohio State facing a team such as Tennessee in the Cheez-it Citrus Bowl.
All of this highlights two interesting aspects of the new college football playoff. First, despite the tripling of the size of the field, there are still plausible scenarios where some very good teams get left out. That said, they generally involve those teams losing at least two games.
Second, the inherent structure of the tournament means that the seeding will not closely correlate to the actual strength of each team. The expected result is likely to be a more chaotic bracket than the madness we experience every March in the NCAA basketball tournament.
To demonstrate this point, Figure 2 below gives the odds for each of the 12 seeds to win the college football playoff in two very different situations.
 |
| Figure 2: Odds for each seed to win the college football playoff using the results from the 2024 simulation and assuming that the teams were seeded based on power rankings. |
The stripped green bars show the results of a previous analysis where the top 12 teams in the nation based on power rankings (which correlate to point spreads) were seeded from top to bottom. In this case, the No. 1 seed has a 21% chance to win the tournament.
The odds fall for the other teams with byes down 14% for the No. 4 seed. There is a sharp drop to 7.5% for the No. 5 seed (because the remaining teams need to play one additional game) and then the odds gradually fall to about 1% for the No. 12 seed.
But the odds from the simulation of the 2024 season follow a much different pattern. The odds for the No. 1 seed start a bit higher at 25%, but they drop to 5% for the No. 4 seed. The odd for the No. 5 seed then jump up to 14% which is the third best in this analysis. The odds then fall again as the seed increases.
Interestingly, the odds for the No. 5 seed are almost as good as the odds for the No. 2 seed. The odds for the No. 6 seed are only slightly worse than the odds for the No. 3 seed, and the odds for the No. 4 seed are actually worse than the average odds for the No. 8 seed. The odds for the No. 9 seed and below are comparable in both scenarios.
If we think about the structure the tournament and use the example from my disruptive scenario as an example, what is going on is clear. In this case, a bye was given to both the Big 12 and ACC Champion, but Utah and Miami were only the No. 7 and No. 8 ranked team in the bracket.
It stands to reason that the first at-large team selected (the No. 5 seed) is most likely the second best team in the SEC or the Big 10. That team is often stronger than the Big 12 or ACC champion. In the disruptive scenario shown above, Mississippi, Penn State, Michigan, and No. 9 seed LSU would all be favored on a neutral field to beat both No. 2 Utah and No. 4 Miami.
When the real college football playoff takes place this winter, expect some upsets and expect some of the betting lines to not correlate with the seeding. It is simply baked into the structure of the tournament.
Seeding Principles, Revisited
With this in mind, when I seeded the teams in the disruptive scenario shown above, I made some choices that I am not sure if the committee would make or not. I moved some of the teams around to avoid rematches and to balance the bracket.
Specifically, No. 5 Mississippi, No. 6 Penn State, and No. 7 Michigan were shuffled to avoid putting the SEC champion (Texas) and runner up (Mississippi) in the same half of the bracket. At the same time, I placed both Penn State and Michigan in the opposite side of the bracket as Big Ten champions Oregon. Potential Big 12 and ACC rematches are also easily avoided.
Furthermore, I intentionally placed Penn State instead of Michigan in Texas' pod due to the non-conference game that the Wolverines and Longhorns will place against each other in Week Two. In my scenario, the two teams do face each other eventually, but not until the semifinals.
In the past, the committee has stated that they will not consider rematches and that they will construct the bracket based simply on their (highly unreliable) ability to properly rank the teams. However, more recent exercises suggest to me that they will do something similar to what I have done above.
In either event, the analysis above suggests that the committee would not be doing the No. 1 seed any favors by placing the strongest at-large team as the No. 5 seed. Placing the strongest at-large team in the bottom half of the bracket creates a more balanced tournament, which is precisely the purpose of seeding the tournament in the first place.
My advice for the Selection Committee is simple: do not get overly hung up on "correctly" seeding teams No. 5 through No. 12. Even if they can correctly rank the teams (which they cannot) the structure of the tournament makes this goal pointless. Instead, the committee should create the most balanced and fun tournament that they can.
That means avoiding rematches and balancing both sides of the bracket based on the actual strength of each team and not just on their resume. Hopefully they will see the light.
Comments
Post a Comment