College Football Has A Sampling Problem

Alabama was ranked #1 in the AP Poll at the conclusion of the 2017 season. This was no surprise as they are regularly near the top of the poll and they had just beaten Georgia 26-23 in the national championship game. They finished the season 13-1, with their sole loss to the Auburn Tigers earlier in the season. But they won the national title, so that loss is redeemed.

Georgia was ranked #2 in the AP Poll. Given that they made the championship final and nearly won the game, this ranking seems fair. They finished the season 13-2 with their only losses being to Alabama in the championship game and Auburn earlier in the season. Interestingly, Auburn has wins against the top two teams in the AP poll. Auburn finished 10th in the poll with a 10-4 record.

Oklahoma, Clemson, and Ohio State fill the 3rd, 4th, and 5th places in the poll respectively, each with identical 12-2 records. Clemson and Oklahoma made the four team playoff and Ohio State missed out having two losses prior to the playoff. Oklahoma and Clemson made it both with a single loss. If you are keeping track of the win-loss records, every team that made the playoff had a loss prior to the playoff.

Taking the 6th spot in the poll is Central Florida. Two years prior they finished the season winless with a record of 0-12. The year prior they finished with a 6-7 losing record. In their final game of the season, they participated in the Peach Bowl with the Auburn Tigers, the team that beat Alabama and Georgia. An SEC Powerhouse. The most interesting fact is that Central Florida finishes the season as the only team without a loss. Beating the team that beat #1 and #2. 13-0. And they finish 6th.

The only team without a loss ... and they finish 6th

Central Florida claimed a national title for the 2017 season despite not participating in the College Football Playoff which Alabama won. Of course, Central Florida was not invited by the College Football Playoff committee despite being the only Division 1 A team without any losses. Without Central Florida experiencing a loss throughout the entire season there isn't a strong case to deny their 2017 championship claim.

In the Final College Football rankings for 2017 which showcased the four playoff teams selected, Central Florida was ranked #12. This put Central Florida below 3 loss Auburn (#7) and just above 4 loss Stanford (#13). You could easily argue that they weren't given a fair shake and their win against Auburn proves that they were certainly underrated at this point in time.

The main argument for their ranking was strength of schedule. Central Florida did not belong to a Power 5 Conference and thus by all metrics they had a weaker schedule than other Top 10 teams who participated in Power 5 conferences. There was also the good old "Eye Test". Basically the committee organizing teams by their own personal criteria of what a "Good Football Team" looks like. Given that the committee is composed of mostly Power 5 athletic directors and former coaches, this criteria can at times be rather subjective.

Now this doesn't mean their exclusion of Central Florida was incorrect. But the criteria that they use to select and organize the teams is rather nebulous and potentially subject to certain biases. This all comes down to a core problem of College Football. Teams simply don't play enough games to differentiate each other in a significant enough way where a team can make it clear through their win-loss record that they at least deserve a shot at a championship.

In college basketball, this problem doesn't exist. If you win all of your games, you are guaranteed a position in the Tournament. Granted college football only has a 4 team playoff and college basketball has a 68 team playoff, college basketball has a format where any team worthy of competing for the championship is given a high seed in the tournament. Evidence of this includes Wichita State appearing as a 1 seed in 2014 despite the fact they weren't in a good conference (they were undefeated though).

Other divisions of college football also have extended playoffs which eliminate this sample size issue in that more teams are included to avoid potential debate and alternative championship claims. Division 1 A has a long history of not having a playoff and only until recently has the championship playoff pool expanded to four.

Some pundits argue that increasing the number of teams decreases the significance of conference championships, but this point seems weak when you see that 2017 national champion Alabama didn't win their conference division and was selected for a spot. Back when the playoff was a championship was a single game, Alabama were also 2011 national champions despite being second in their division. So winning a conference hasn't really meant that much given prior precedent.

There's also the notion that the best team should be labeled the championship despite the results of the game which is more likely to occur when more teams are allowed in the playoff. This argument is against fluke winners, but any tournament champion would need to fluke win multiple games which kind of invalidates this argument.

But let's say you want to protect against fluke champions and preserve the importance of conference championships and keep the playoff at four teams. You need an objective criteria that gives all teams in eligible pool to have a chance to earn a spot in the championship. The notion that some teams have zero shot of winning a championship based on factors outside their control is ridiculous. Given that the sample size is small, wins and losses alone are not sufficient enough in that you could theoretically have as many as 11 or 12 undefeated teams. And there is some credence to strength of schedule be a criteria to differentiate similar teams.

But whatever criteria is selected, a team needs to be able to overcome their schedule by some means. The most obvious metric is the margin of victory which may rub people in the wrong way (due to a dislike of "running up the score"). A lot of computer models use margin of victory in their models because it gives them a meaningful way to measure performance against other teams outside of their win-loss ratio and their schedule. It is also an element of a game a team can actually control unlike their schedule.

So there are two solutions to the current issue of college football committees ignoring teams from weaker conferences in terms of selecting teams for a national title opportunity. The easy option is simply to expand the field and guarantee that undefeated teams are included in the playoff. The other option is to eliminate the committee as a selection body and use an objective criteria (ranking model) which measures at least some metrics that an individual team can control (like margin of victory).

Why should anybody care about how a sport decides its champions? On the surface it doesn't really matter and is rather pointless, but sports are at least more exciting and enjoyable when your local or favorite team has a path to a championship that the team itself can control rather than needing other teams to lose to become eligible. A game isn't a good one if a path to winning that game isn't one within a player's or team's reach.