Open it from the Prospects page (How it works, below the rookie board) or the button above. It starts with the Seer grade: what it is (a rookie’s chance of a top season from his draft slot and age, then his rookie year after 8 games), how well it did (every draft class from 2008 forecast only from the classes before it: how well it ranked hits above misses against the draft slot alone, and a table of “when it said 20%, it happened 27%”), and what was tested and left out because it made those forecasts worse (combine numbers, the landing spot, order within the position, boosted trees).
Below that is the research, position by position (pick one with the tabs): the takeaways and the hit rate by draft range are open, and “Show all the research” opens the other nine sections. It studies every QB, RB, WR and TE who entered the NFL from 2000 on and has played three seasons: combine invitees, draft picks and undrafted players who made it. A hit is a top-24 PPR finish (top 12 at QB and TE) in one of his first three seasons.
Rookie seasons: the clearest signal after the draft. Players whose rookie year was in the top third for points per game hit far more often in seasons 2–4, even compared with players drafted in the same range (only rookies with 6+ games count, and rookie production partly reflects playing time). Younger rookies tended to do better too, but those gaps are small enough to be chance. The recent rookie classes are listed with how rookies who had a first year like theirs did next.
The trait table splits the position into thirds on each measurement and compares the best third with the worst. The key column, "Same draft range", repeats that comparison only among players picked in a similar range (picks 1–10, 11–32, 33–64, 65–120, 121+, undrafted), so it shows roughly what a trait added beyond where teams drafted him. "Priced in" means a big gap overall that disappears once draft range is equal. "Still matters" means the better third kept hitting more often, and it holds up allowing for the 14 traits tested per position; "Suggestive" clears an ordinary 95% range but could be chance.
The headline numbers say how well each kind of information told hits from misses on players the model didn’t see (0.5 is a coin flip): the draft slot is by far the strongest, the build alone is useful before the draft, and together they add little over the draft slot.
What the stars had in common: every player with at least one top-12 season (QB and TE top-6), against everyone measured the same way: how many came from the better third of each measurement, the stars who came from the worse third (the exceptions), and their draft rounds. It describes the stars without allowing for draft slot; the trait table does that.
Built like the stars: the measurements that clearly set a position’s stars apart (age everywhere; speed score and the 40 at RB and WR; height-adjusted speed at WR and TE; at TE nearly every drill), as a checklist with the cut-offs. The table shows how players who matched it did against what their draft slots predicted, and the honest answer is: barely better, because NFL teams already draft on these traits. Recent rookies are listed by how many they meet, with the past stars they’re built most like.
New measurements: height-adjusted speed (speed score × height, for WRs and TEs), BMI, and our own athletic score: a player’s average percentile at his position across the drills he ran (40, vertical, broad, 3-cone, shuttle; three or more), 0 to 10. It isn’t Relative Athletic Score, which is someone else’s measure.
Build a profile: enter any mix of age, height, weight, 40 time, jumps, agility drills, rookie-year points per game and draft range, and see every past player like that, with the hit rate, its 95% range and the rate their draft slots alone would have predicted (so you can see whether the profile beat the draft). With a rookie-year PPG the outcome is a hit in seasons 2–4, after the rookie year. Recent classes shows the last three draft classes with their percentiles, the build-only chance, how their draft range has done, and the closest historical builds.
Data: nflverse (combine and draft data courtesy of Pro Football Reference), seasons through last year. Hand size and arm length aren’t included (no licensed source yet), nor is college production. It’s history, not a forecast.