College Football Analytics: Playoff Contender Tiers
Ranking the 0-1 loss teams with advanced stats
It is hard to believe we are already at the midway point in the 2021 college football season. Every team has played almost half their schedule, and we’re starting to get some questions answered about certain programs. The problem is, it seems like for every question answered, 10 more pop up. As is the nature of our beloved college football!
This season will be the 8th time a committee has selected 4 teams to compete in a playoff for the national championship. Much debate has gone into not only the size of the playoff, but the criteria in which to select the teams for the playoff. Record, strength of schedule, conference championships all seem to be the first things that pop in our head when we rank teams, and you would think the committee would focus on these core tools. While they do take those into account, we also have to hear the committee talk about less clear things like strength of resume (Is that a feature on Linked Premium?) and “Game Control” (Say that in a bar in Tallahassee in 2014 and you were certainly met with an alcoholic clothes on shower).
Here is how the committee has selected its participants so far:
The two things that stand out: No two loss teams, and unfortunately no Group of 5 teams. In order to look for our possible participants in 2021, we need to take both of these into account. There are currently 33 teams with 0-1 loss. Our goal is to put these 33 teams into tiers in order to look at the College Football Playoff landscape.
So how do we do it? I can think of no better way than using our trusty K-Means Clustering algorithm. For those of you who are new to the newsletter, the algorithm is friend of the program, as I have used it to put teams into clusters entering the 2021 season, and I have used it to put QB prospects into groups heading into last years NFL Draft. The 5 second explanation of K-Means Clustering is feed algorithm data, it groups teams together by how close they are to each other based on the data we provide.
The data we will be using is a mixture of rating systems. First, we will be using my own personal power ratings. Next, we will use each team’s win percentage combined with their expected win percentage based on post game win expectancy. Finally, we will be using Brian Fremeau's FEI ratings and Bill Connelly's SP+ ratings to get a complete mixture of advanced rating systems. Let’s press the magic button and get some tiers!
CFB Playoff Tiers
Outside of Clemson, the three teams in the first tier are the teams you would have expected to be here coming into the season. Alabama is recently coming off a loss to TAMU, and Ohio State suffered their lone loss at the hands of Oregon. As much as we might want to see a playoff without Alabama, they are still firmly in control of their destiny in the SEC West, and therefore the playoff.
A lot of talk around these programs in the first tier seems to center on how big of a gap there is between them and the rest of college football. This gap can actually been seen in this clustering exercise when you plot out the clusters through principal component analysis (PCA, aka smush into 2 dimension so we can plot on x/y):
As you can see there is a bit of a gap between the first tier and any other tier. This agrees with the conventional wisdom of college football. This isn’t to say the top tier is unbeatable, but rather you should expect at least two of them in the playoff. How to sure up the gap is a conversation for another day.
In order to analyze the makeup of each tier and why each team is in that tier, we have to look at the underlaying data and make inferences based on what we find. Taking the average of each data point used can give us a great resource for analysis.
*Note* I couldn’t fit all of the variables used in this table, so I stuck with the core essentials
Right off the bat we can again see the divide between the first tier and everyone else. They lead in power rating, expected win percentage, and every advanced metric on this table.
Tier 2 Analysis
Moving away from the first tier, the second tier consists of teams who could find themselves in the playoffs, but still have some questions lingering about their team. We know how elite Iowa has been on defense, but their offense is currently ranked 75th in offensive F+, which is 5th worst among these 0-1 loss teams. Iowa may very well defense their way into the playoff, but if the defense happens to give up some untimely explosive plays, can the offense step up? In this tier 4 Big Ten teams (Michigan, Michigan State, Iowa, Penn State), all of which have to play some combination of each other. It will be interesting to see whether they cannibalize each other, or will someone emerge and give Ohio State a run for their money.
Outside of Cincinnati (G5 breakthrough), Notre Dame (offensive issues), and Oklahoma (offense/QB controversy?), the rest of these teams seem to be here simply as a “we dare you to finish with 1 or less loss”. Kentucky has a MASSIVE game looming vs Georgia, and Oklahoma State still has the meat of their Big 12 schedule to showcase their abilities.
Tier 3 Analysis
The third tier is reserved for Group of 5 teams with one loss, or teams that will most likely get shafted by the committee regardless of their performance on the field. The obvious team sticking out is the undefeated Wake Forest Demon Deacons. So why is Wake Forest in this group? The answer primarily lies in the “wins over expectation” category. Overall the tier is first by a long shot in that category, and Wake Forest has the highest wins over expectation of any 0-1 loss team. Wins against Syracuse (13.9% post game win expectancy) and Louisville (12.9%) show some narrow escapes for the Demon Deacons.
Coastal Carolina boasts one of the best offenses in college football, is one of the most entertaining teams to watch, and has a good chance to finish the regular season undefeated. Will the committee show them love and give them a playoff spot? 0% chance. Their Sun Belt schedule will never be enough for the committee, so they have been effectively eliminated before the season began. It seems like a system that auto eliminates teams before the season is a bad system, but that is a discussion for another day.
Tier 4 Analysis
The fourth tier is the tier loaded with teams you should keep an eye on as the second half of the season kicks off. NC State’s lone loss to Mississippi State might be enough to keep them out of the playoff, but they do control their own destiny in the ACC Atlantic after knocking off Clemson. Wins against Boston College, Wake Forest and Pitt in a potential ACC championship would be good enough for a NY6 Bowl at the very least.
The same can be said for Baylor in the Big 12, and Arizona State and Oregon in the Pac-12. At the moment these teams are fringe top 25 teams according to my power rating, FEI and SP+. When you look at wins over expectation, they’re right around where they should be given how they have played. However, given their current standings any of these teams will shoot up the rankings if they win out, and force the committee to take a good look at their body of work.
Tier 5 Analysis
The final tier unfortunately will not be making it to the college football playoff. These are group of 5 teams that just don’t have enough juice to force a conversation from the committee. However, these teams are having seasons that should bring great pride to the fans of these programs.
Houston’s defense ranks 29th in the country in Defensive F+, and the Cougars are currently atop the AAC. Houston recently accepted a bid to the Big 12, and things are looking up for them on the football field. Sounds like the Cougars are living the good life!
UTEP has 3 seasons with 5+ wins since 2010, and this season will be their 4th. UTSA has made precisely 2 bowl games in their time in the FBS, but they are well on their way to their 3rd appearance. Wyoming might be 0-1 in conference play so far, but they’re 4-0 outside of their conference and they have Josh Allen to root for on Sundays. These programs deserve their time in the limelight for their seasons thus far, and if they continue to win games hopefully they are put in bowl games that give them the ability to showcase what their program offers.
Conclusion
This is where we stand so far this season. Given the overlapping conferences seen with these teams, we will end up losing some teams to multiple losses. Unless the college football committee does something its never done before and puts in a 2 loss team, 4 of these teams will be playing for a chance at the national championship. Clustering gives us the ability to group these teams together, and make observations based on groups and the gaps between these groupings. I hope you all enjoyed this little exercise and hope you’re ready for the second half of the season to begin!
If you want to dive in to the data like I do, check out @CFB_Data and @cfbfastR on Twitter, where you can learn how to get started in the world of College Football data analysis!
If you want to see more charts and one off analysis, follow my twitter page, @CFBNumbers