College Football Mathematical Preview: The Playoffs and NY6

So far this summer I have performed an in-depth, data-driven analysis of Michigan State University's football schedule, the Big Ten race, the rest of the Power Five, the FBS independents, and Group of Five. Today, it is time to put all of those pieces together and to make some predictions as to what teams will make the Playoffs as well as the other four New Year's Six Bowl Games.

You Wanna Talk About Playoffs?

In the previous parts of this series, I have presented data tables for every conference that included playoff odds for all 130 FBS teams. Table 1 below shows the 25 teams with the best odds to make the playoffs along with each team's preseason tanking, strength of schedule (with the national ranking in parenthesis) and the odds for those teams make the final game and to win the National Championship. 

Note that my strength of schedule calculation is based on the number of expected wins for an average Power Five with that schedule and therefore a lower number means a harder schedule.

Table 1: Top 25 playoff, final game, and national title odds based on the results of a 100,000 cycle Monte Carlo simulation

In general, the playoff odds mirror the preseason rankings fairly closely with a few notable exceptions. The top five teams appear in order, with the exception of Alabama. The defending champions are the consensus No. 1 team in the preseason, but the Crimson Tide grade out with only the fourth best odds to make the playoffs (24 percent) behind Ohio State (27 percent), Oklahoma (30 percent), and Clemson (39 percent).

The reason for this is essentially the difference in the strengths of schedule, where Bama has one of the top 10 most difficult schedules, while the remaining challengers have schedules outside of top 30. Clemson in particular has one of the easiest schedules in the entire Power Five (ranked No. 67 overall) which helps to explain the large difference in playoff odds between the Tigers and the rest of the field.

That said, the National Title odds for the top four teams all range from between nine and 12 percent per team, and Alabama's odds leapfrog Ohio State's odds. In other words, if Alabama can survive the SEC gauntlet, their odds go up since, on paper, they are the best team in the country.

For the rest of the Top 25, deviations between the preseason rank and the playoffs odds rankings are usually due to differences in of strength of schedule. The main exception to this rule is Notre Dame. 

The Fighting Irish have one of top 25 most difficult schedules, but their playoff odds ranking (No. 6) is noticeably higher than their preseason ranking (No. 11). This seems to be purely due to the fact that as an independent, they do not have to play in a Conference Championship. 

As we learned during Big Ten basketball season, the best way to ensure that a team does not lose is simply not to play. This "strategy" appears to boost Notre Dame's playoff odds.

Just looking at Table 1 suggests that Clemson, Oklahoma, Ohio State, and Alabama are most likely quartet of teams to make the playoffs this year. Based on the raw odds, that is essentially true. However, when I crunch the numbers on different combinations of those top four in the playoffs, I get the result shown in Figure 1.

Figure 1: Distribution of the different combinations of the preseason top four teams in the Playoffs based on the Monte Carlo simulation. (Note that Clemson is labeled simply as "C")

According to this figure, if my calculated odds are correct, it is likely that only one or two teams from the group including Clemson, Oklahoma, Ohio State, and Alabama will make the playoffs. These data also suggest that none of the those four teams making the playoffs is a more likely scenario than three or all four members of the group making it.

On one hand, this is not surprising. My results say that Clemson makes the playoffs in almost 40 percent of the simulations. However, this also implies that in 60 percent of the simulations the Tigers do not make the playoffs. So Figure 1 does seem to follow from the numbers in Table 1. 

In addition, this distribution reminds me a bit of distribution of No. 1 seeds that make the Final Four in March Madness. Both simulation and history tells us that only one or two top seeds usually advance to the final weekend.

But, football these days is a different animal. First of all, the odds that I give for the top teams to make the playoffs are low compared to the odds that can be derived from the current Las Vegas money line (which give the top four teams between a 60 and an 80 percent chance to reach the Playoffs). 

Furthermore, a retrospective analysis of the past seven years suggests that between two and three of the teams that start the season in the top four usually end the season in the playoffs. I would expect (and in fact project) that trend to continue in 2021.

What this highlights is a potential weakness in my simulation methodology: I treat all teams equally. I assume that all teams are equally likely to be both better than expected and worse than expected. In reality, however, teams like Clemson and Alabama, at least over the past few years, rarely underachieve. They consistently "beat the odds" which is why they are so dominant. That seems likely to continue this year. That said, these trends are often only trends until they aren't.

As a final piece of data related to the big picture aspects of the playoffs, Figure 2 give the odds that at least one team from each conference will make the playoffs (on the left) and the odds that more than one team from each league will be invited (on the right). 

Figure 2: Distribution of the numbers of teams from each conference in the Playoffs, based on the results of the 100,000 cycle Monte Carlo simulation

In a nutshell, the SEC is most likely to get at least one and also multiple teams into the playoffs, which is not a surprise. After that the ACC and the Big Ten have the best odds. Interestingly, the Big 12 has better odds than the Pac 12 to get one team into the Playoffs, while the Pac-12 has better odds to get multiple teams. 

I suspect that this is due to the fact that the Pac-12 has divisions and the Big 12 does not. A team like Oregon could be edged out of the division race by Washington and still make the Playoffs. In the Big 12, however, the team that loses in the conference championship game is most likely out of the Playoffs and the third place team's odds are likely longer than the odds of division runner up from another conference.

Disruptive Playoff Prediction

With the data shown above, I could simply make my play-off predictions by taking teams with the best four odds and then slotting the next eight or so teams in the New Year's Six Bowls. Based on the average outcome of my 100,000 simulations, this is perhaps what is most likely.

However, the reality of the situation is that eventual Playoff and NY6 participants are not selected based on averages. They are selected based on the outcome of one very specific scenario: the actual results of a full season of football. In order to make a more interesting projection of the 2021 post-season, I instead turn to my "disruptive" simulation.

As I summarized in the previous installments of this series, my disruptive simulation tweaks the paraments in order to create a historically reasonable number of total upsets. In my preview of each conference, I summarized the expected records for all teams assuming that the favored teams all win and if I "disrupt" the simulation.

Table 2 gives a summary of the division and conference races in both scenarios.

Table 2: Summary of the division and conference races base on the raw odds and if the simulation is disrupted with the insertion of upsets.

This table is divided into two parts. The left side shows the odds-based favorite in each division, which is the equivalent to the favorite if the preseason rankings are correct and if there are no upsets. Basically, the highest ranked team in the preseason always wins. The odds for each team to win the division are shown and the team favored to win the conference championship game is shaded in green.

The right hand side gives the results of each conference race in the disruptive simulation. I have also included the projected record for each team following conference championship game in this scenario, as well as key wins and losses. The divisions where there is a change in the winner between the two simulations are highlighted in yellow.

In addition to the teams that the disruptive simulation projects to play for a conference championship, Table 3 summarizes the notable remaining power five teams with two or fewer losses overall.

Table 3: Projected records and notable wins and losses for other play-off and NY6 contenders based on the disruptive simulation

For the purpose of my "official" playoff projections, I will use the more interesting results from disruptive simulation.

If the season were to play in this manner, Clemson would emerge as the lone undefeated team and would obviously be given the No. 1 seed in the playoffs. Ohio State, at 12-1 with several quality wins would be the obvious option as the No. 2 seed, but the other two slots are less straightforward.

In this scenario, Texas A&M is a two-loss SEC champion with some high quality wins. I think that the Aggies would earn the No. 3 seed. Then, it would most likely come down to Big 12 Champion Oklahoma, at 12-1 or Notre Dame at 11-1. I think that the committee would have to give the nod to Oklahoma in this case as the No. 4 seed with the message that winning a conference title trumps Notre Dame's accomplishments. (Sorry, Irish, but if you don't like it, go join a conference.)

So based on the current bowl rotation, the Playoffs would set up as follows in this scenario (with the projected point spreads shown):

• Orange Bowl: No. 1 Clemson (-1) versus No. 4 Oklahoma
• Cotton Bowl: No. 2 Ohio State (-2) versus No. 3 Texas A&M

In this scenario, Clemson would be expected to face Ohio State (+1.5) in the National Title game, where the Tigers would be a narrow favorite.

As for the other New Year's Six Bowls, based on the projected standings and season results summarized above, I would expect the following match-ups (with the projected spreads shown):

• Rose Bowl: Wisconsin (-4) versus USC
• Sugar Bowl: Alabama (-3) versus Iowa State
• Peach Bowl: Notre Dame (-1.5) versus Florida 
• Fiesta Bowl: Cincinnati (-6.5) versus Utah 

I could also see Alabama and Florida trading spots, but my pairing produces a likely more competitive set of games (based on the preseason rankings) and also avoids an Alabama / Notre Dame rematch from last year's semifinal. 

As for the Fiesta Bowl, Utah makes a surprise appearance as the last remaining one-loss team. I pair the Utes with the best Group of Five team in Cincinnati, who would edge out Boise State and UL Lafayette with a slightly better resume.

Note that there are a few teams that I do no project to secure a New Years' Six slot but who would project to be a tough out in a bowl game, including Georgia, Oregon, and North Carolina. That trio are all consensus top ten teams in the preseason, but they all have tricky schedules.

So, that is what the math says about the 2021 college football season. While this is all well and good, it would be nice to know if all of this math that I have done might actually be useful in the betting market. After all, my simulation projects odds that can be directly compared to many of the same odds for various season outcomes as set by our friends in Las Vegas.

In the final installment of my preseason college football analysis, I will dig into these number in more detail to provide a preseason dose of Bad Betting Advice. Stay tuned and Go Green.