I betcha the AP is only marginally more sensible.
The CFP rankings are probably the best because they seem to be more of the mindset of the NCAA selection committee - i.e. thinking in terms of "resume" - but, as we all know, even they are flawed.
I will get virtual bricks thrown at my head for saying this, but I would select something like the top (i.e. most accurate over a ~10 yr period) 3-5 predictive computer models and base the CFP seeding purely on that.
Two things here:
1. If you, like me, played the ever living crap out of the NCAA Football games back in the day, you know that the games did a pretty credible job of simulating the behavior of the real life "Top 25" polls. They did it with code of course and I think it would be very enlightening to see the formula(s) used, as they would have had to take into account the kinds of (IMO) silliness we as humans tend to apply to where college football teams should be ranked. Most notably, team prestige and the idea that teams can't/shouldn't be moved up or down too quickly. About the only thing you can say for this is that it essentially creates "smoothing" in the rankings, which creates a perception of credibility relative to the computer models, which themselves fluctuate relatively wildly throughout the open weeks of the season, but then calm down into very sensible (and demonstrably accurate) assessments as we move into the back half of the season. But even the CFP (wisely, IMO) doesn't release their polls until we are some ways into the season, because the CFP committee will tell you the same thing the people who code computer models will tell you: we don't have enough data yet to make these even worth looking at yet.
2. Going further, I think we like to keep humans involved in these rankings for the wrong reasons. The CFP committee may look at various computer models in making their decisions but, ultimately, it's a human poll. IMO, far and away the #1 reason, the
real reason that we keep humans in the CFP loop: to protect the blue bloods and to make sure we keep the TV ratings up (which is of course closely related to "protect the blue bloods"). My point has been emphatically proven this year, because we have 1-loss Notre Dame team that, IMO, should be nowhere near the CFP...but they have something very special on their resume that only humans can see: they are a 1-loss Notre Dame team! Just for context, look at the end of season Sagarin top 10 (bolded are the CFP teams):
| Alabama | 104.77 | 11-0 |
| Ohio State | 97.37 | 6-0 |
| Clemson | 96.98 | 10-1 |
| Oklahoma | 92.01 | 8-2 |
| Georgia | 91.91 | 7-2 |
| Florida | 90.51 | 8-3 |
| Notre Dame | 88.88 | 10-1 |
| Texas A&M | 88.54 | 8-1 |
| Iowa | 87.49 | 6-2 |
| Iowa State | 86.51 | 8-3 |
If, for argument's sake, we had simply let the Sagarin model select the top 4, that fourth spot would go to Oklahoma (by the skin of their teeth). To get to Notre Dame, you have to skip over Oklahoma, Georgia, AND Florida.
When all is said and done, in practical terms, the value of the CFP this year is going to be to make Alabama "prove it". In other words, if Alabama wants to be National Champion, they are going to have to beat 2 very good football teams. If they do that, we outsiders may not necessarily be happy, but we'll at least sleep well knowing we are pretty sure we know who the best team in the land was this year. That said though, a CFP berth is worth an awful lot in terms of both monetary and prestige terms and I hate to see the system being used to help the rich get richer (Notre Dame). We have a serious "zero or 1 loss brand name team" fetish in college football that has screwed us (and will continue to screw us) out of a lot of good matchups over the years.
