College Football Analysis: Fixing the 12-Team Playoff

In early June, A sub-group of the College Football Playoff's management committee presented a proposal to expand the current four-team College Football Playoff into a 12-team tournament. Over the last week or so, a few details have started to emerge. Here are the key points of this proposed expansion:

1. The six highest ranked conference champions would automatically qualify for the tournament
2. The four highest ranked conference champions would be seeded No. 1 through No. 4 and given a first round bye.
3. The remaining two conference champions and six additional at-large teams would be seeded No. 5 through No. 12 and would play in a series of four first-round games on the campus of the higher seed at some point in mid-December
4. The quarterfinal round would be played on or near January 1st, and the report suggests that teams would be assigned to their traditional bowls in the quarterfinal rounds.
5. The semifinals and finals would be played at some point in January, but the bowls or stadiums that will be used for these rounds are not clear.

I recently conducted an analysis and criticism of the 12-team format. Based on my analysis, the odds for a team ranked outside of the top eight teams has less than a five percent chance of winning the National Title, suggesting that such a Cinderalla story would only happen about once every 20 years. 

While I am sure than any team out there finishing the regular season ranked No. 10 would take those odds, when I consider the additional wear-and-tear on the college athletes (remember when that was a thing?) and the threat that the 12-team play-off poses for the traditional Bowl system, for me, as a fan, this is a bad idea. As I have stated previously, "Eight is enough."

That said, a 12-team play-off seams all but inevitable. With that in mind, I have a few suggestions on how to make it the best 12-team playoff possible.

What's Wrong With the Current Proposal

With a 12-team tournament, the format proposed by the sub-group in which a total of four teams get byes is essentially the only way to structure the event. The fact that the bye seem to be reserved only for conference champions seems fine to me. Upon reflections, this actually builds in some nice flexibility that I think will come in handy.

One implication of this rule is that teams that fail to win their division titles, yet still wind up ranked in the top four (such as Ohio State in 2016 and Alabama in 2017) will not be given a bye. It also means that Notre Dame, as an Independent, will not be eligible for a bye.

The other sneaky part of the proposal is the idea that the top six conference champions will receive automatic bids. First, this means that at least one Group of Five team will always make the field. But, it does not guarantee that all of the Power Five Champions will also make it. In fact, based on the final playoff standing in 2020, Pac-12 champ No. 25 Oregon would have missed out on the tournament in favor of No. 12 Coastal Carolina. This fact obviously did not escape the notice of the Pac-12 commissioner.

I have no specific concern with these aspect of the proposal. What I do have a major issue with is the following statement from the initial press release from the sub-group:

"The playoff bracket would follow the rankings, with no modifications made to avoid rematches of teams that may have played during the regular-season or are from the same conference."

This may seem innocuous or even as a positive to some, but I would argue that this is, by far, the weakest part of the entire proposal. It will make preserving traditional bowl match-ups, such as the annual Big 10 - Pac-12 clash in the Rose Bowl almost impossible, and as we will see, there is absolutely no justification from a competitive view point to be this strict with seeding. 

In order to see this problems more concretely, Figure 1 below shows the likely playoff brackets that would have been generated following the seasons from 2014 to 2020 if the proposed 12-team playoff were in place. Note that the number before each team corresponds to their final ranking in the College Football playoff poll.

Figure 1: Likely 12-team playoff brackets for the seasons from 2014-2020 had the currently proposed 12-team playoff been in place.

I see several problems with these brackets, many of which are highlighted. The teams shown in yellow are all involved or potentially involved in a conference rematch either in the first round or in the quarterfinal round. I count a total of 14 potential rematches in just seven years. 

The potential to see an average of two early round rematches in the playoffs per year should be enough to turn any fans stomach. Around 25 percent of the first eight games of the playoffs project to be rematches in this format. That is, frankly, unacceptable.

As for traditional bowl match-ups, giving the top four conference champions bye will make this task much easier, with the expectation of the traditional Rose Bowl match-up. If we assume that the higher ranked conference champion out of the Big Ten or Pac-12 will be given the Rose Bowl in the quarterfinal round, a traditional Rose Bowl match-up was only possible in 2014, 2015, and 2016, less than half of time. This is also unacceptable.

How to Fix It

Fortunately, a few simple tweaks can be made to the pairing that would essentially eliminate rematches and greatly increase the odds of a traditional Rose Bowl match-up. Figure 2 below shows the changes that I would propose.

Figure 2: Proposed changes to the 12-team brackets from 2014-2020 to avoid rematches and preserve the Rose Bowl match-up

In this case, the teams shaded in yellow have changed their position on the bracket to avoid rematches or in an attempt to preserve a potential Rose Bowl match-up. As we can see, in seven-out-of-seven years, all potential conference matches up through the quarterfinal round have been eliminated, and it was not that difficult to achieve. This rule change is an absolute must in any final 12-team format.

Protecting the traditional Rose Bowl match-up is more difficult. In 2020, the Pac-12 does not even make the tournament, and Ohio State would have drawn the winner of the Texas A&M - Coastal Carolina game in my proposed bracket. From 2014 to 2016, and in 2019, the traditional Rose Bowl match-up would have occurred, assuming no upsets in the first round. In 2017 and 2018, the Pac-12 champion would have needed to win in upset fashion to force the traditional match-up. 

In the 12-team format, pairing the actual champion of the Big Ten and Pac-12 looks to be difficult in the quarterfinal round. Basically, of the two teams would have to finish as the fifth or sixth ranked conference champion, or both teams would receive a first round bye (as was the case in 2014 or 2016). But, in those cases there was a non-champion from the Big Ten (Michigan State and Ohio State) who were available to place in the "Rose Bowl pod" of the tournament. In most years, I would expect this to be the case.

A possible criticism of this plan is that deviating from the strict rankings of the teams would somehow cause a competitive imbalance in the tournament, somehow making it less fair. However, as I showed in detail previously, this is simply not true. Making small adjustments to the placement of teams will have a negligible impact on the eventual National Champion.

To demonstrate, I performed a series a calculations of each of the proposed brackets in Figure 1 and 2 using estimated and consistent power ranking to generate win probabilities for each potential game. The average change in National Title odds for any given team between the proposed and my "optimized" 12-team tournaments was less than half of a percent.  The greatest deviation was only two percent. 

The year that showed the biggest deviations total between the proposed and my alternative bracket was in the year 2016. Figure 3 below shows the results of the calculated National Title odds in the two scenarios. As we can see, the difference is truly negligible.

Figure 3: Comparison of the calculated National Titles odds for the 12-team bracket proposed by the sub-group and for a revised bracket that protects the Rose Bowl match-up and avoid rematches

The amount of uncertainty in the true strength of each team will make a much bigger impact. In the example above, the No. 1 ranked team would be favored to beat the No. 2 ranked team by about 3.5 points. If I artificially make the No. 1 team a little bit better, such that they are instead favored by 5.5 points, the No. 1 seed's odds improve by six percentage points, which is a much larger impact than any change to the bracket itself.

New Bracketing Principles

In going through exercise of creating the revised brackets, a new set of bracketing principles become apparent. First, the fact that the sub-group is already seeding the top four champions No. 1 to No. 4 and giving them byes makes the placement of these teams into bowls trivial. Based on tradition, the obvious choices are:

• Rose Bowl: highest ranked champion from the Big Ten or Pac-12
• Sugar Bowl: SEC Champion
• Orange Bowl: ACC Champion
• Cotton Bowl: Big 12 Champion
• Peach Bowl: Big Ten Champion (if the Pac-12 Champion is placed in the Rose Bowl and the Big 10 Champion is also ranked in the Top four) or Group of Five Champion, based on geography
• Fiesta Bowl: Pac-12 Champion (if the Big Ten Champion is placed in the Rose Bowl and the Pac-12 Champion is also ranked in the Top four) or Group of Five Champion, based on geography

This scheme clearly gives a bit more preference to the top four bowls in this list, but that was already the case historically prior to the current playoff and New Years Six structure. The Peach Bowl and Fiesta Bowl only appear a total of twice seven year in my proposed brackets. However, upsets in conference championship games, which have been rare over the last few years, would likely result a higher Peach Bowl and Fiesta Bowl quarterfinal contests. It would also help if the Pac-12 were more competitive.

As for the bracket, the No. 1 seed would line up to face the lowest ranked other team with a bye in a potential semifinal contest. The other two teams would be paired on the other side of the bracket.

For the first round games and position, the selection committee would then prioritize placing a Big Ten or Pac-12 team into the same pod as the Rose Bowl quarterfinal. There may be a rare situation where both the Pac-12 and Big Ten champion are ranked outside of the top four of conference champions. In this case, the Rose Bowl would not participate in the playoff and would instead invite the highest ranked remaining Big Ten and Pac-12 non-champions.

After that, the committee would assign teams to the bracket based on the rankings in using an "s-curve" philosophy. However, the committee should have not just the option, but the directive to specifically avoid rematches, to the extend possible. As shown above, this would create a much more interesting and fun tournament.

As I have also stated previously, I would take the additional step of giving the Selection Committee the task of setting up compelling match-up to enhance and potential save the existing bowl system. Current conference bowl tie-ins are not necessary. The Bowl games should simply bid for a certain space in the hierarchy, and the committee would then slot teams into bowls based on this hierarchy. 

For example, the following match-ups could have been created in 2020, if the committee had this power:

• Peach Bowl: No. 13 North Carolina versus No. 14 Northwestern
• Fiesta Bowl: No. 15 Iowa versus No. 16 BYU
• Citrus Bowl: No. 17 USC versus No. 18 Miami
• Alamo Bowl: No. 19 Louisiana versus No. 20 Texas
• Sun Bowl: No. 21 Oklahoma State versus No. 22 San Jose State
• Outback Bowl: No. 23 NC State versus No. 24 Tulsa 

As a fan, this looks like fun, and I feel this is likely the only way to save the tradition of the bowls.

But... Eight is Still Enough

While College Football seems to be hurtling towards a 12-team tournament, I still feel that an eight team tournament is adequate. In my previous explanation, I proposed a series of eight-team tournament brackets. However, I think that I leaned a bit too hard into the current NY6 bowl parings. In reality, the SEC has a strong historical connection to the Sugar Bowl, but the Big 12 has more history with the Cotton Bowl. 

With this in mind, I have revised my proposed eight-team tournament bracket using essentially the same bracketing principles that I outlined above. I would take the top six conference champions and two at-large teams. The Big Ten and Pac-12 Champions are automatically paired in a quarterfinal game in the Rose Bowl, and the other top three seeds are given preference to their traditional bowl, with Notre Dame given preference to the Peach Bowl. My revised brackets for the 2014-2020 tournaments are shown below in Figure 3.

Figure 3: Revised eight-team brackets for the 2014-2020 seasons using my proposed bracketing principles.

At this point this proposal is essentially in the "old man shouts at clouds" realm, but as an old-school college football fan, this looks as close to ideal as possible. It won't happen, but man, would it be fun.