I might someday actually devise a program to rank teams in some way that I find favorable, but until them, I'm going to lay out the premises behind what I consider the most useful determinants for ranking.
1. Record. Clearly, winners should be ranked above losers.
2. However, winners who beat other winners should be ranked above winners who beat losers, and losers who closely lose to winners should be ranked about the same as winners who closely beat losers. And this should go at least three-deep into the schedule: opponents record, opponents' opponents record, and opponents' opponents' opponents record. This should also, in theory, give enough closure to the system to take into account approximately 150 teams but only 12 games.
3. Division I Playoff Subdivision games should be counted differently than Division I Bowl Subdivision games. Whether that means they only count as .75% of a win (and 1.33% of a loss), or a team is penalized for playing a Division I Playoff Subdivision team matters less than taking this into account.
3A. Similarly, playing within your conference allocation should count differently than playing outside of it. This would be better served by including both information on the conference—playing and closely losing a BCS conference game counts more than playing and handily winning a non-BCS conference game—and by accounting for Athletic Department budget—schools with budgets of over $75 million (Texas) who play schools with budgets of $13 million (North Texas) are clearly not operating on a level playing field.
4. Home field advantage counts. Home wins should count less than road wins; perhaps 8 home wins is equal to 6 road wins. Neutral wins aren't as good as road wins, but are better than home wins.
5. Score counts, but not raw score. We'll do it logarithmically. That is, if a one-point win results in the winner being considered as being one point better than the loser, a 10-point win makes the winner only two points better than the loser, while a 100-point win makes the winner three points better than the loser. (Clearly, there are some bugs to be worked out).
6. Because football is a game played by very young men, the emotional aspects of the game should count, too. Therefor, both the rankings at the time of play are taken into account—beating the AP #2 team in the country in the second week of the season takes effort, even if that team ends up losing five or six games afterward—and the rankings at the end of the season—beating the team that eventually becomes #2 at the end of the season in the first game of the year still counts, even if they weren't ranked there then—are important to acknowledge.
7. The rankings shall not be released until there is sufficient data to make the rankings comprehensible. At the bare minimum, this would require four or five games.
I have no idea how to compile these general guidelines into a comprehensive system, however. But it's a start.