The Wolverines beat Western Michigan 13–12 on a Saturday night in Ann Arbor. They survived. Their standing in this poll did not: 14th in our preseason edition, 47th now. Clemson lost by 41 to LSU and fell fewer places.

A win is a win. Apparently, a resume is another matter.

Before we get carried away, one important distinction: in August, we asked who would finish the season with the strongest resume. This time, we asked what teams had actually accomplished. Those are different questions and the answers changed sharply. The people participating can change between editions, too.

AP voters also dropped Michigan out of their top 25. The difference here is the size of the reshuffling, not that one poll noticed and the other did not. AP voters submit an ordered top 25. We ask a narrower question: among these four teams, who has the strongest and weakest resume based on games already played?

This edition draws on 155 eligible ballots. It is a snapshot of those participating fans’ judgments within our fixed 50-team field.

What the voters moved toward

LSU is the clearest case. Eighth in the preseason edition, first now, after a 51–10 dismantling of Clemson. Miami climbed from seventh to second after Malachi Toney’ing Stanford. Ole Miss went from 13th to third on the strength of Lucas Carneiro’s leg. Virginia, 39th in August, is 10th here and 25th in the AP poll - its offensive line is scary and should have PHDs by now.

Those are not small adjustments around a preseason pecking order. They are a different opening argument about who has accomplished the most.

And to the Missouri and Texas Tech fanbases looking at this chart: yes, we saw the wins. But an unbeaten record does not tell the whole story. As the schedule develops, fans will have more evidence to weigh. That can move teams up… or down.

Losing well beat winning cheap

At least in some of this week’s comparisons.

Boise State lost to Oregon and rose seven places from our preseason edition. Baylor lost to Auburn and rose one. A loss did not automatically push a team down this ranking; just as a win did not guarantee a climb.

Opponent quality looks like part of the story. But the poll measures the judgments, not the reasons behind each vote. We cannot say where Boise State or Baylor would have finished had they played somebody else, and while there is no automatic FCS penalty built into the model… teams with a FBS opponents seemingly received a head start.

The preseason favorites offer another striking contrast. Ohio State went from first to fifth, Oregon from second to eighth, Georgia from third to seventh, Indiana from fourth to ninth, and Texas from fifth to 11th. That does not prove reputation stopped mattering. It does show that the August order was not simply carried forward.

The part that should make everyone humble

The top of the chart is less concentrated than it was in August. This week’s top four account for 44% of the total index across the field; LSU accounts for 20%. Those are shares of an index - not shares of votes, playoff probabilities, or a measure of how certain each fan feels.

The same caution applies to comparing scores across weeks. Each edition averages 100 across the same 50 teams, but it remains its own snapshot. A large jump in the index is not a precisely measured multiple of improvement.

You will also notice a broad second tier. The model found four tiers across all 50 teams, but only the first two appear in the top-30 chart. Finer breaks did not clear tier rule. Sharing a tier does not mean teams have identical resumes, and 100 as the field average is not a tier boundary.

One weekend gives us results. It does not give us certainty. LSU has the strongest opening argument but it rests on one Saturday night in Death Valley. I don’t think anyone of us would be surprised after this Saturday night’s tilt of Ohio State and Texas that we might see someone new on the throne, or a return of the king. And, that is why we keep asking the question. And why we love college football.

How our FBS Resume MaxDiff poll works

Each voter saw 18 unique four-team groups and two separated repeat screens, choosing the strongest and weakest year-to-date resume in each group. Balanced forms distributed comparisons across the field, with randomized card order. Voters could skip unfamiliar groups instead of guessing.

After the poll closed, eligible choices were fit with the frozen symmetric MaxDiff model. Uncertainty and tier stability were estimated by resampling whole voters 2,000 times. Familiarity skips and repeat screens were not treated as additional preference choices.