At the center of the biggest controversy at the end the of the 2017 College Football Season was the University of Central Florida (UCF), coached by second year head coach Scott Frost. UCF, which had finished with an abysmal 0-12 record in 2015, ended the 2017 regular season at 12-0 and champions of the American Athletic Conference. When College Football Playoff (CFP) selections were announced UCF alone stood undefeated yet ranked only 12th in the CFP selection committee poll.
UCF was not selected to participate in the CFP despite finishing the regular season undefeated. The controversy created by choosing four other participants with one loss over UCF began almost immediately, and was not eased when UCF defeated Auburn in the Peach Bowl 34-27. Auburn was responsible for both losses by teams in the CFP Championship Game, having defeated both Georgia and Alabama during the regular season.
This article presents a simple statistical algorithm to evaluate the strength of play of college football teams and, in so doing, attempts to provide a way to objectively answer the question of whether UCF should have been included in the 2017 CFP. It will not address the philosophical question of whether any team, regardless of conference, which finishes its regular season undefeated, should be granted an automatic bid to the CFP. That is a question unanswerable by analytics and should remain the sole domain of internet message boards and sports talk radio.
The Only Statistics that Matter
The list of statistics used to describe college football is long. The data set I used for this analysis had more than 120 different metrics baked in, and that doesn’t include any calculated statistics like yards per play or completion percentage. These statistics are useful, but at the most basic level, there are only two statistics that determine whether a team wins or loses: How many points the team scores and how many points the opponent scores.
One way to measure how well a team scores points is to compare the number of points a drive results in to the expected number of points for a drive beginning at any point on the field. A drive can end in 7, 3, 0, or -2 points. (The actual number of missed point-after attempts is small enough to ignore in the computations.) Figure 2 shows the expected points for drives beginning x yards from the end zone since the 2008 season.
Drive Point Efficiency
Drive-Point Efficiency is a statistic that measures the average number of points scored per drive compared to the expected points of those drives. Figure 3 shows a typical offensive quarter for a team.
The team scored 17 points on 5 drives. Based on the starting spots of those drives, the expected points were 14.9.
DPE = (17-14.9) / 5 or .42.
Offensive and defensive DPE are calculated the same way. Scale this up to 13 drives per game on offense and defense, and 12-14 games per year, and we have a solid metric for determining how efficiently a team converted scoring opportunities into points and prevented their opponent from scoring points.
DPE does an excellent job of describing how efficiently a team maximizes its opportunities to score. But it is only half of the story. The other half is how many chances a team gets to be efficient. That is measured by average number of offensive and defensive drives per game.
Consider Navy. Since 2008 Navy has been one of the most efficient teams on offense. Only Oregon at .90 was more efficient than Navy at .83. But Navy was also dead last in average number of drives per game. Navy had an average of 11.8 drive opportunities on offense per game. The FBS average was 13.8 per game. If Navy were just average in number of drives it would have an extra 1.7 points per game to work with on offense every game.
The graphic below illustrates the relationship between offensive DPE and average number of offensive drives per game. Teams in the upper right quadrant were strongest at both from 2008-2017. The group of top performers is not a surprise. Oregon, Oklahoma, Oklahoma State, Baylor, Texas Tech, and Houston have consistently produced offensively for the last decade.
In a similar vein, the graphic below shows Defensive DPE and average defensive drives per game. Note: because negative values on DPE and lower values for average defensive drives per game are better, both axes are inverted. This keeps the elite teams in the upper right quadrant. As should surprise no one, Alabama was head and shoulders better on defense that the rest of the country from 2008-2017.
Combining offensive and defensive DPE and average number of offensive and defensive drives provides a single number that incorporates how dominant a team is on both offense and defense.
The Top 10 combined most efficient teams were, from 1 to 10, Alabama, Boise State, Ohio State, Oklahoma, Oregon, Wisconsin, TCU, Florida State, Clemson, and Oklahoma State.
The chart below plots the two parts of the Efficiency Advantage formula. As before, the best teams since 2008 are in the upper right quadrant.
Where do CFP Teams Fall?
Generally, the teams that comprise the Final Four of the CFP are above average for offensive efficiency and the best of the best on for defensive efficiency. The exceptions to this were Oklahoma in 2017 and Oregon in 2014. There appears to have been a shift in the four years towards a strong bias for defensively superior teams.
A more granular look at 2017 looks similar to the CFP playoff poll. The participants in the 2017 CFP are highlighted.
Similarly, 2016, 2015 and 2014’s CFP participants are on the upper right as well.
You might recall the controversy surrounding the 2014 CFP selections. Because of the Big 12’s decision not to have a conference championship game, neither Baylor nor TCU were able to make up the points to edge out Ohio State in the rankings. Florida State remained in the Top 4 because it was undefeated and coming off a national championship in 2013. Nevertheless, Florida State’s record belied what was a win-by-the-skin-of-your-teeth approach to the season, and their efficiency advantage showed that. Florida State’s efficiency advantage in 2014 was the worst of any CFP participant since. Oregon exposed Florida State in the semi-final round by a score of 59-20.
Should UCF have been the College Football Playoff?
UCF’s strongest argument was that it was undefeated. All it had to do was point to 2014 Florida State to make the argument that an undefeated team, however strong their on-field performance, deserved the right to play for a championship. 2014 Florida State, however, was riding a 27 game winning streak. UCF won 6 games in 2016 and 0 games in 2015, so UCF didn’t quite have the benefit of the doubt argument that Florida State got.
Looking at their on-field performance in 2017 a case can be made for both sides of this issue.
UCF sits along the leading edge of the 2017 teams and is right in line with the other CFP teams. UCF’s offense was almost as efficient as Oklahoma’s and significantly more efficient that Georgia’s, Alabama’s, or Clemson’s. It isn’t, however, in line with bias towards defensively superior teams (unless your team starts with “O” of course).
If the intention were to put the four best teams in the tournament, then it should have been Clemson, Alabama, Ohio State, and Wisconsin. Note, Wisconsin is hiding behind Clemson.
If the intention were to put the four most deserving teams in the tournament, then Ohio State and Oklahoma are certainly out, having been pantsed in Iowa City and at home in Norman by Iowa State. (Kudos to the state of Iowa for really messing up the CFP picture.) UCF is in based on winning all the games it played.
The other teams are all equally undeserving. Alabama lost the Iron Bowl and didn’t even play in its conference championship game. Georgia lost to Auburn. Clemson lost to Pitt but beat Auburn. Wisconsin lost to Ohio State in the conference championship game. Penn State didn’t even play in the conference championship game.
UCF was undefeated and the CFP selection committee had precedent in including a manifestly sub-par yet undefeated team. UCF did not have the Power 5 conference pedigree, and there is statistical evidence to demonstrate that the worst Power 5 conference played better football, on average, than the best Group of Five conference. (stay tuned for that article, m’kay?) UCF’s strength was its offense in 2017 but its defense was barely better than average and the general trend is to favor defensively superior teams. Still, Oklahoma was selected with a less efficient defense.
In the end, I’m going to punt on making a decision. I think a compelling argument can be made both for and against UCF’s inclusion in the 2017 College Football Playoff.
What do you think? Should they have been in or out?