One of our longest "traditions" is preseason polls, and discussions thereof.
Last year, 11 out of 25 ended up unranked.
So, our "experts" are right half the time, slightly better than.
Cincy, Dawg, my man.
What does this mean? If you went to a newspaper reporter or whatever SID/gofer votes for coaches and said "Are you an expert in predicting final polls in college football." They would all answer in the negative. The point of the preseason poll is not universally to say "This is how it looks at the end."
To do such a thing super well would require modeling out a ton of games. Last year bowls, or matchups that don't even exist, dropped four teams out and added four more. If you found a person who could hit at even 65 percent on an average annual basis, that person would not be a gofer or sportswriter, but a wealthy gambler or bookmaker.
It also speaks to the fact that polls must serve all masters. They are asked to predict, but also to reflect what has happened thus far this season. Each week they must reflect last week's results, and the season overall, and carry us toward and end. Sometimes they're asked to reflect resumes, other times how good a team is. In the end, nothing can do that competently.
What's more, this opens two more problems. The first is that if your goal is to have a more correct final poll, you're better off with Texas and ND rather than Cincinnati or Army. Those teams were ranked last year, but on average, stacking a list with historical powers is better to boost that out of 25 number. Beyond that, even if you were the genius who had army 19th, Kentucky 12th and Cincinnati 24th, you still need the rest of the voters to agree with you in groupthink, or the wisdom of the crowd wipes it away with conventional thinking.
In the end this is in some ways about trying to tear down and sneer at "experts" whose pretend expertise is endowed on them by us. They are for the most part unlikely to say they are great predictors, yet we pretend they do just so we can say "no you're not."
(Secretly the best way to rank week-to-week is to take advanced stat rankings and tweak for wins and losses to your liking. Far more satisfying)