Monday, January 26

A Better Consensus

Prior to 1992 there was no reason to expect any consensus as to who the best team was unless one team was head and shoulders above the competition. The national champion was determined by two independent polls who periodically came to different conclusions.

In 1992 the first feeble attempts at pairing the top 2 teams in a bowl game were began, becoming the BCS in 1998 and adjusting to its current form in 2006.

Even with this system two sources of uncertainty remain that have proven capable of casting a shadow of doubt over the identity of the true champion. Often the claim to the #2 spot is hotly contested. If the #2 team upsets #1 in a close game and #3 pounds the #4 or #5 team in their bowl who should be #1? If a team goes undefeated with a weaker schedule due to historical ties most concede they likely are not the best team, but how can one ever know for sure? And what if they win big against a perceived giant?

In the former case the consensus is reduced by a close score in the ratings. In the second case it is reduced by a significantly increased variance of the opinions of the untested team.

Close ratings can be mitigated by increasing the distance between teams at the cutoff for national championship consideration. An impressive win might be enough to close a large gap between two adjacent teams but it would not make up for the additional increase one of the teams ahead of them would get for their win. They would be shooting for a moving target and one can only score so many points.

The variance in opinion of an undefeated team can be mitigated by giving them an appropriate test. Undefeated teams usually end up at least #15. With a win over a top 5 team this would be increased to near #5. Another would surely get them into the top 3. But with two wins over top 5 teams in a championship environment and ending unscathed, plus losses by #1 and #2, who would be left to make a claim against the only unbeaten team? Two games is enough.

Include all undefeated teams.
Include all teams above the first gap of 1.5 in the rankings average.


This alone does not guarantee that an undefeated team will be adequately tested or that there will be more than one team.

If the first gap is between #1 and #2 use the second gap in the rankings average.

These criteria directly address the only significant sources of doubt created by the current system, significantly increasing the consensus of the national champion while keeping the field to a very limited number.

The polls with no formal playoff was used as the championship system for 56 years, making this the longest championship system used without structural modification at any level of American football.

The second longest was that employed by the NFL from 1933 to 1967. That system involved one championship game between the division champions with extra games added if they were needed to break a divisional tie. Such a contingent structure is very well suited to accommodate the two criteria above.

No comments:

Post a Comment