Model Score: how the number is built
Model Score is one number, 0 to 100, for every college player: how good is he at what he is supposed to be doing, where he is doing it. The median is about 70. A 95 is rare.
What it is not
It is not a prediction of the NFL player he becomes. That is the whole point, and it is worth being blunt about, because the number will sometimes look wrong if you expect otherwise.
Diego Pavia rates 93.2 and went undrafted. Anthony Richardson rates 73.9 and went fourth overall. Both are right. Pavia was genuinely excellent at playing college quarterback; NFL teams were buying something else. Richardson was drafted on arm and size, not on what he actually did at Florida. A number that measures college production is supposed to say so.
If you want a draft projection, the big board consensus rank is the column for that. Model Score sits next to it on purpose - the interesting players are the ones where the two disagree.
What goes into it
Five ingredients, weighted. Everything is measured against players at his own position, so a guard is compared to guards.
- His best season (50%). The single year he played his best football. This is the biggest piece because it is the clearest evidence of what a player is capable of.
- His current season (20%). What he is doing right now. This is what lets the number fall, which a best-season-only rating can never do.
- His career average (20%). Whether the good year was the pattern or the exception.
- When he broke out (5%). Doing it as a freshman counts for more than doing it as a fifth-year senior. We use how long he had been on campus, not his age - age quietly leaks how the NFL treated him, which would make the score partly a draft projection again.
- Size for the position (5%). A small piece, and deliberately small. Traits are what the draft over-weights; this is a nudge, not a thumb.
Freshmen with no college tape yet are carried by their recruiting profile, capped so that pedigree alone can never produce an elite score. A player resting on one season carries a note saying so.
Why it moves
It is recomputed every week. A senior having a worse year than his junior year goes down, and should. Of the college players with more than one scored season, a little over half finished lower than their peak. That is the number behaving correctly, not drifting.
Where it is honest about being weak
Cornerback is the hard one. The best corners get thrown at least, so their box-score numbers look like a bad player's. Interior offensive line and tackle disagree with where those players actually get drafted more than any other position, which we read as the draft valuing traits there rather than the model being broken - but we cannot prove that without All-America and award voting data we do not have yet.
We publish all eleven positions anyway. A number that only appears where it looks good is not a number you can trust.
Coverage
14,983 college players have a Model Score across eleven positions. 9,027 of them have a profile page - those are the players with two or more scored seasons. A player with one season has a score that leans mostly on his recruiting profile, and a page built on that is worse than no page.
Where to see it
- Big Board - a sortable Model column next to the consensus rank
- Prospect Analytics - sorted by Model Score
- College Team HQ - every roster, plus Position Rooms ranked by Model Score
- Any college player profile page