Why the College Football Playoff Is Terrible But Better Than Before

The sample size is still embarrassingly small, but at least there’s less room for the availability cascade.

On January 1, the University of Oregon and Florida State will play in the Rose Bowl while the University of Alabama and Ohio State will face off in the Sugar Bowl. The winners will meet at AT&T Stadium in Arlington, Texas, on January 12 to contest the first formal national championship in Division 1 college football. The new format is an exciting change, a struggle between the four best teams in the land to be crowned the true national champion. It has inspired pundits and casual fans to enjoy countless hours of debate about the merits of individual teams vying for a spot in the championship game.

The new system is also a terrible method for deciding the strongest college football team in Division 1.

The availability cascade is a fancy name for the concept that voters are susceptible to narratives rather than actual results.

Playoffs, by their very nature, are a flawed system. They rely on a sample size of one deciding game, an all-or-nothing affair unrelated to what came before it. Part of the magic of football (and sports in general)–the any given Sundayness of the game, the fact that an inferior squad will triumph against a better one on an individual basis with relative frequency–invalidates the playoff system as an arbiter for superiority. There’s too much luck and randomness involved in sports to ensure that the best team wins on an individual game basis. This season’s college football national champion won’t necessarily be the best team; they will simply be a good squad that played better in two games a full four months after the season began. Yes, the four teams in the playoff had strong seasons–they wouldn’t be there if they didn’t–but there’s no guarantee the team that prevails will be the best among the quartet.

And yet, the playoff is still a superior method to the previous one. That version, the BCS rankings, pitted the first- and second-ranked team against each other in a final game. The teams came from a complex formula that equally weighted a series of computer rankings and two separate human-driven polls, which were the problem. Human voters are notoriously unreliable, easily susceptible to a number of psychological factors that affect their judgments without them knowing it.

One of the major ones is confirmation bias, which played a major role in the media poll. Here’s what ESPN BCS analyst Brad Edwards toldPsychology Today:

Most media members are asked to make a national championship prediction before the season starts and, if they are voters, they’ll fight to keep that prediction alive for as long as it can be justified. In other words, if there are several undefeated teams or several one-loss teams, and one or two teams must be separated from that bunch at the end of the season, many voters will ignore what their eyes see and instead try to fulfill their preseason prophecy by elevating their team of choice.

Anchoring, in which someone relies on a single piece of information too much when making a decision, was another factor. Consider a coach whose team got badly beaten by another team. If he had a vote in the coaches’ poll, he might rate that team more highly than it deserved because his first-hand experience made such a strong impression.

Finally, the availability cascade is a fancy name for the concept that voters are susceptible to narratives rather than actual results. If a team gains a reputation as a title contender early in the season, it’s very hard to disabuse voters of that notion. All these factors made the human polls unreliable at best and downright wrong at worst.

Of course, the College Football Playoff Selection Committee can be affected by some of the very same biases, but they don’t release their first set of rankings until the last week in October, which gives them time to form opinions independently of the later media avalanche. It’s not a perfect system, but at least it’s less subject to the availability cascade.

The fairest model for crowning a champion is what many European soccer leagues use, a single-table format where every team plays every other team home and away, and the team with the most points at the end of the season wins. In some years that model sacrifices a bit of drama (and always loses out on revenue generated by playoff games, which is why some leagues are turning toward playoffs), but the larger sample size means the best team more frequently emerges on top. Consider this: Pretend a superior team has a 55 percent chance of winning any individual game. That’s not a great advantage over the course of one game, but it becomes quite dramatic over the course of a longer season.

That is obviously an untenable model considering the number of college football teams, but conference structure mimics some of this behavior. Of the five power conferences, only the Big 12 selects its champion by the winner of a regular season round robin in which every team plays every other team. (The other conferences have a one-off championship game between the two divisional winners.) This year, Baylor and TCU shared the title after each posted an 8-1 record in the Big 12 and went 11-1 overall. They finished just out of the playoff, fifth and sixth, respectively. No one said sports were fair.

Related Posts