2 prospects · 2018 draft
Baker Mayfield and Lamar Jackson project within 8 picks of each other, and neither bust risk nor the historical comps can separate them. Take the one at the position of greater need.
The four numbers a GM checks first.
Each axis is the prospect's percentile among prospects at his position (0–100). Bigger shape = stronger across the board.
Top five historical comps per prospect. Shared comps highlighted.
Top three concerns and top two upside signals from the model.
Projected share of his team's snaps across the first four NFL seasons. Solid = seasons already played, dashed = the forecast.
The model's plain-English read.
Draftanomics projects Baker Mayfield (age 23.0 at draft) as a top-10 pick at #3 (range #1-#60, very wide band — too little on him yet to project tightly). Bust probability sits at 0% — low. The comp set is strong: 4 of 4 closest comps with known NFL outcomes became starters or better. Closest comps: Kevin Hogan (Starter), Paxton Lynch (Starter), Brandon Allen. Ceiling is Marcus Mariota (Franchise); floor is Brandon Allen. Age 23.0 at draft. Slightly older than the cohort median (~22.5); upside ceiling is trimmed. The call is a hold at slot: the model's #3 sits within 12 spots of the room's #15 — no market edge to chase, so take him where he sits, and let the 0% bust probability break ties on draft day.
Draftanomics projects Lamar Jackson (age 21.3 at draft) as a Day 2 pick at #60 (range #19-#101, very wide band — too little on him yet to project tightly). Bust probability sits at 0% — low. The comp set is strong: 4 of 5 closest comps with known NFL outcomes became starters or better. Closest comps: Jared Goff (Starter), Paxton Lynch (Starter), Dak Prescott (Starter). Ceiling is Jared Goff (Starter); floor is Kevin Hogan (Starter). No Combine drills or Pro Day numbers on file while most of the class has tested — the athletic part of the grade is unmeasured. The call is a genuine fade: the room's consensus prices him around #29 and the model won't take him before #60 — a 31-spot markdown. The 31-spot markdown rests on thin verified signal rather than a named flaw — the model hasn't seen enough yet to match the room's conviction.