Two scores, kept apart on purpose. The UIR is the Upside Index Rating: his sustained NFL starting-role potential, rather than elite NFL quality. The PROJ Upside Draft Projection is where he is likely to be drafted. A board that blends the two cannot tell you whether a player is ranked for his production or for his draft stock; this one shows both.
Who is on the board. Players who can enter this draft under the NFL's rule (three years out of high school) and still have college eligibility, read from PFF's eligibility year for every player. Underclassmen who are not yet eligible are on the watchlist classes instead.
UIR (0-100). The chance he becomes an NFL starter, predicted from his college profile alone: PFF-graded production and usage in every stat family (passing, rushing, receiving, blocking, pass rush, coverage, run defense), his previous season and career, his share of his team's output, his grades measured against players at his position and level, his team's quality that season, whether he plays in FBS and a Power Four conference, how many players his school had drafted before that season, his years in college and his position. A starter means 32+ games with at least half of his unit's snaps in the five NFL seasons after his final college season, drafted or not. The consensus board, the draft and scouting opinions are not inputs. The scale is fixed: 90 means about a 60% chance, 70 about 25%, 45 about 5%, in every class.
Upside Draft Projection. A projected pick and round, a likely range, and the chance of being drafted and of going in round 1. In the fall it comes from the college profile and the consensus big board, blended at the weight that predicted real drafts best when it was learned on past classes with a genuine pre-draft board and tested on a class it had not seen. Past classes were projected the same way, never using their own draft slot.
How both were tested. Historical ratings exclude each player's own NFL outcome from his fitted model. These are retrospective reconstructions: ordinary forward tests can use training outcomes that matured later, and the earliest class uses leave-season-out training on later classes. Model selection has reused these cohorts; strict outcome-maturity tests are reported separately. The numbers, the formulas and the tests that failed are on the methodology page and in the past-class view.
Tiers. The top two are earned, not allotted: Blue Chip needs a UIR of 93.5 or more inside his class's top 32 (in past classes 85% of those prospects became starters and 56% made a Pro Bowl). Generational requires a UIR of 97 or more inside the class top 32, all three position signature traits at or above the 92.5th percentile, and at least 10 games and 400 snaps in the signature season. There is no quota. This is a conservative prospect designation, not a calibrated prediction of an NFL generational career. The stricter rule reduces mature historical labels from 14 to two; both became multi-Pro-Bowl stars, but the sample is only two (95% interval 34% to 100%) and reuses classes involved in signature selection. It is not an independent accuracy improvement. College trait percentiles are not opponent-adjusted; the UIR separately uses school and competition context. Every player card shows the traits and sample counts. Everyone else is placed by a draft-round band of his UIR rank: Round 1 (1-32), Day 2 - Early (33-64), Day 2 - Late (65-105), Day 3 - Early (106-180), Day 3 - Late (181-260), then Priority Free Agent, Developmental and the watchlists. Below the top two, a tier is a rank band, not a rating threshold.
Star odds. Beside the chance he becomes a starter, each prospect carries the share of past prospects with his UIR who made a Pro Bowl or an All-Pro team. Models trained on Pro Bowls directly were tested and ranked future stars worse than the UIR does, so the odds are read off the UIR itself, measured on the 2017-2021 classes.
During the season. A player's current season is blended with his previous full season by how much of it he has played (5 of 12 games: 42% current), the rule that tested best on the first five weeks of past seasons.
Where it is weakest. The UIR cannot see film, medicals, the combine or interviews, which is why the real draft still beats it at the top of a class. The projection leans on the consensus board, so a player the board has not ranked yet is projected from his college profile alone. Early in a season the current year is a small sample.