Scorecard
How we actually did
Every ranking published here gets scored against what actually happened, and against the expert consensus captured before kickoff. Nothing is left out — a week where the model was beaten appears the same way a week where it won does.
How to read it
Starters we got right is the one that matters. Take the twelve quarterbacks who actually scored the most that week: how many of them were in our top twelve? "7 of 12" means seven were, and five of the players we told you to start were not top-twelve that week. That is the honest quality of any weekly list, ours included.
Points captured is the same idea in points: what our starters scored as a share of what the best possible starters scored. 100% would mean our lineup was the perfect lineup.
"With Vegas" and "model only." Since week 1 of 2026 the published rankings lean heavily on Vegas lines. The model-only version is kept and scored on the same games, so you can see every week whether Vegas earned its place. If it stops, it gets dialed back.
Order is rank correlation across the whole list — 1.00 is a perfect ordering, 0.00 is a hat. It is what the research literature uses, so it is here for comparison; a leading expert consensus has been measured at about 0.35 for a weekly quarterback list.
No week has been scored yet. The first scorecard appears once week 1 is complete and the box scores are final — an honest one has to wait for the Monday night game like everybody else.