Brock Purdy

I raised him to QB7 from QB19 and he lands QB3 on value. He finished QB23 last year, so be honest about what this is — a bet on the situation, not on what he has done.

GAMES ONLY, corrected 22 Aug. An earlier version of this row today also claimed a RATE of 18.30/g, and that claim was WRONG and is withdrawn. It argued the model credited him with only 86% of San Francisco's pass plays while Bo Nix took essentially all of Denver's. That compared pass ATTEMPTS against pass PLAYS, which are different things -- plays include sacks and scrambles. Measured over 95 team-seasons 2023-25, a team's primary quarterback throws a MEDIAN 0.808 of his team's pass plays, p90 0.878, and Purdy's own realised seasons were 0.808 and 0.799. His 0.865 was ABOVE the median and near the 90th percentile, and his projected 28.36 attempts sat on his career 29.57. There was no dilution and nothing to fix. WHAT SURVIVES IS THE GAMES, which is independent of that error. The model had 12.6, the largest haircut of any startable quarterback, off ONE injury-shortened season; his own outlook line already flagged it, and the measurement behind the flag is that QB availability explains just 9% of next-year availability. 13.3 is his own 2023-25 record, 40 of 51 (16, 15, 9) -- his actual durability rather than his worst year. THE RATE IS LEFT NATIVE at the model's 15.90. It does sit 3.82 below his own decay-weighted 19.72, the largest negative in the QB group, and that is worth understanding -- but traced, it is ordinary regression of his EFFICIENCY, not a volume bug: his yards per attempt has fallen 9.64, 8.49, 7.63 across three seasons and the model lands 7.65, right on his most recent. Whether it over-regresses a passer's efficiency is a real question and is NOT settled here, so no number is invented for it.

This is a view, not a correction — the model said something different and was overruled. See how it works.

Pace extrapolates the selected weeks over 17 games. It is what that stretch was worth, not a projection — no availability, matchup or regression adjustment is applied.