@Alderian - While individual, apparently random factors certainly can distort a result of a game (we'll take the WSU-Colorado game as an example here), the nature of randomness is such that a team most likely will not get through an entire season, or even half a season, surviving on mere luck. A few lucky games, even in a row, perhaps, but not six, or seven, or nine or ten. I do wait until we're about halfway through the season for this reason [and for another - by now we actually have some idea of how valuable a team's wins and losses are].
And I know that head-to-head comparisons against one mutual opponent don't mean anything (if they did then Oregon would probably lead Oregon State), but taken over a lot of mutual opponents (division rivals) they do, because having a lot of mutual opponents means that their strength of schedule is roughly equal, at which point it would make sense to say "team F beat teams K-P by more points than team G, team F has done more than team G, team F will be ranked higher than team G." Remembering that the rankings are based not on head-to-head matchup strength but instead what they've accomplished, this is internally consistent and usable.