Evaluator compass
How the boards lean
Every draft board from 2020 to 2025, placed by how it ranks prospects against the field. Centre court is consensus. Hover a dot to see the evaluator, click or tap to pin their card under the map, and scroll or pinch to zoom in.
Pinned evaluators
Click any dot to keep its card here. Pin a few to compare them.
How to read it
The axes
Each board is placed by the kinds of prospects it ranks above or below the field. Right means it bets on tools and pedigree (); left means it bets on production (). Up means it moves size up (); down means it moves self-creators up (). Units are standard deviations across boards. Past about one SD, distance from the centre is drawn compressed so the outliers fit; the rings mark true 1, 2, 3 and 4 SD.
The card
Grade is the v12 grade per pick and its rank among every graded board. The signature prospect is the pick that shows the board's lean most clearly: a prospect it ranked well above the field, of the type its position on the map predicts, weighted toward the top of the draft. Best and worst calls compare the GRBU value implied by the board's rank and by the field's median rank with what the player has produced. A positive number means the board landed closer than the field did.
What it can't tell you
Position on the map says how a board disagrees, not whether it's right. That's what the grade is for. The two axes explain about a tenth of how boards differ from the field on rankings they weren't fit to. Most disagreement is about individual prospects. Boards with few rankings sit closer to the centre because their lean is less certain, and 2025 outcomes are still early.
Method
For each ranking we take the board's log rank minus the average log rank of every other human board for that prospect, centred within each board. Prospects need at least 15 human rankings. Each ranking is weighted by the site's draft-weight curve, averaged at the board's rank and the field's rank, so a call at #3 counts far more than one at #40. The two axes are fit to those deviations from prospect profiles (age, size for position, role, level, and per-70 and efficiency stats), plus a small prospect-specific term. The fit uses ridge penalties and was chosen by five-fold cross-validation on held-out rankings. Model, mixed and post-draft boards are projected onto the human axes and don't shape them. Grades come from the v12 leaderboard.
Browse every evaluator as a list
| Evaluator | Type | Grade rank | Tools ↔ production | Size ↔ creation | Signature |
|---|