2026 Draft Season
Fantasy Football Draft Rankings
Ranked by our in-house 2026 season model (v2), last updated Sat, 18 Jul 2026 15:49:59 GMT. Every board below sorts by value over positional baseline, not raw projected points — see "What are VBD draft rankings?" below.
Weekly projections
One hub per week, Week 1 through Week 18: projections with ranges once each slate posts, start/sit calls, and grades on the accuracy ledger when the games are final.
Quick answer
What are VBD draft rankings?
Our draft order sorts by value over positional baseline (VBD) — a player's P50 season projection minus the P50 projection of the last startable player at that position (QB12, RB30, WR36, TE12) — not raw projected points. That's why a high-scoring QB can rank below a lower-scoring RB or WR: quarterbacks are drafted from a deep pool, so the 13th-best QB is usually still available; the 31st-best RB usually isn't.
Every projection behind these rankings is a range, not a guarantee — each player carries a P10–P90 band, and real seasons land across that whole range.
FAQ
How often are these rankings updated?
Daily model runs are planned for the 2026 season. Right now, rankings update on-demand as we retrain the model — the "updated" timestamp on each scoring page shows the last run.
What scoring formats are covered?
PPR (1 point per reception), half-PPR (0.5 points per reception), and standard (0 points per reception) — pick the board that matches your league's settings.
Why does a QB with more projected points sometimes rank lower than a RB or WR?
Overall rank sorts by value over positional baseline (VBD), not raw points. QBs are drafted from a deep, replacement-level-rich pool, so even a big point total often trails a scarcer RB or WR in value. Position rank ("QB3", "RB12") is unaffected — VBD only changes the all-position order.
Where's the accuracy record?
It's live on the accuracy page, and it's empty. The ledger opens at zero at Week 1 of 2026, and no week gets written in after the fact. From there, one row a week posts how far our projections landed from real results, averaged over the players we graded. The backtest is already public on the methodology page: four held-out seasons against a repeat-last-year baseline, three won, one lost. The loss was wider than any of the wins.