What's new

The board, team hubs, movers and position pages lead with a dated, written read — what changed, who moved, and why it matters — refreshed with every model run. The table is still there; it just doesn't go first anymore.

Once a month the model publishes its own mock draft and freezes it — no edits, no retcons. Every pick carries the simulator's conviction at that slot, and every pick gets graded against the real draft in April. The July edition is live now.

Prospects with real signal behind them now carry a full written report — the verdict, the case for, the case against, sourced against the boards we track. Thin profiles wait until the data earns the prose; nobody gets filler.

Who really holds the No. 1 pick next April? The 2027 order page now tracks every team's slot odds — seeded from last season's standings today, taken over by real results once the games start in September.

Every team hub's target board is now floored at draftable tiers and weighted by actual roster need — no more kickers padding a contender's list. If a name is on a team's board, the model can defend the fit.

A prospect page that doesn't exist returns a real 404 instead of a blank page pretending otherwise — with a pointer to the player you probably meant. Transfers who changed schools now land on their current page.

Drafttek nailed Saquon, Burrow, Ja'Marr and Kyler. Tankathon and Sporting News round out the top three. Every analyst we ingest gets a public hindsight grade — sort skill, top-32 hit rate, bust avoidance — with their best and worst calls on the record.

Every comprehensive analyst board, weighted by 11 years of hindsight, with the model laid over the top. Verdict pills flag where Draftanomics parts ways with the room — model bet, model dissent, split room. The dossier shows you which writer had him where.

We score YouTube channels on trust (PFF College, ESPN College Football, JD PicKell and the rest) and pull the three best reels per prospect. Jeremiah Smith, Ellis Robinson, Dylan Stewart, Arch Manning — clips on every top-30 dossier.

Seeded from the 2025 standings — LV at No. 1, Tennessee No. 2, Cleveland No. 3 — until the real order lands. Run it, swap picks, trade up, watch the board fall. The model takes BPA-with-need from there.

Seven teams' active general managers updated: NYJ (Mougey), TEN (Borgonzi), JAX (Gladstone), LV (Spytek), MIA (Sullivan), ATL (Cunningham), and MIN (Brzezinski, interim) — keeping the draft record attributed to the people who actually made the picks.

Every team's offensive philosophy, defensive base, coverage shell, and edge archetype manually verified against current head-coach and coordinator hires.

Tabbed prospect dossier, team hub, class summary, and track record. Breadcrumbs everywhere. Research drawer (⌘+/) surfaces methods + glossary contextually.

An audit found eight inputs that quietly leaked post-draft information into training — the model was grading with answers it shouldn't have had. We removed them. Its accuracy dropped, honestly, and the honest number is the one we publish. Receipts on /track-record.

Eleven position-specific bust models, each shipped with a confidence range and a plain label for how accurate that position's model has been historically — so you know which calls to trust and which to take with judgment.

A new signal that combines how far the model sits from the analyst boards with how much those boards disagree among themselves — surfacing the prospects whose evaluations genuinely split the room.