When trying to come up with a system to rank teams, you have to consider what factors of a team's performance is important and how to quantify that. It is not possible to quantify all things, but finding a way to quantify as much as possible as accurately as possible makes a rating system that can be respected. Thinking about rating systems in this way can help relieve the animosity some fans feel towards computer ranking systems. When I made a rating system, I decided early on not to consider the human polls. I find that at the bottom of the ratings are teams which really don't belong; they are included because of popularity more than on-field achievements. If this exists at the bottom, I think in varying degrees it exists throughout. The things I did include: 1) Number of losses 2) A team's performance in comparison to their opponent's opponents 3) Quality of a team's opponents When I first came up with my formula, I did check my rankings against the opinion polls to arrive in the approximate neighborhood of popular opinion. I also used the opinions of other fans who reacted to the rankings I made public. When comparing a human opinion poll to a computer poll, I like to stress the idea that computers follow much more rigidly a set of guidlines. A person may have an initial set of beliefs of what makes a good team, but that set of beliefs quickly becomes clouded and not as strictly followed because of feelings, opinions, and other external influences. That is a good thing in some cases, because there are some people whose intuition guides them correctly. That intuition makes good coaches, players, professionals, and presidents. But I think we can all agree that good intuition is hard to find. How many great coaches, players, professionals, and presidents are there? They are rare, the exception rather than the rule. An opinion poll is the amalgamation of many opinions, and while some of those opinions may be guided by exceptional intuition, many of those opinions fall short of ideal. A computer poll will have it's flaws, but one flaw will not be that it didn't follow a preconceived idea of what makes teams good. The stronger that preconceived notion is, the stronger the poll. In comparison to human opinion polls, a strongly conceived computer poll is no less relevant, and an exceptionally conceived one may in fact closer approach perfection.
IMO computer and human rankings are silly no matter how much thought or analysis goes into it. The only way to say who is better is to settle it on the field.
Which is fine if you are the NFL and have only a handful of teams. Unfortunately there are 119 teams in 1a football.
