The 2026 NFL Season, By the Numbers

This page is the front door for the 2026 coverage here: everything below is computed from the published schedule and 27 seasons of the public game log — no roster projections, no voted rankings, no invented numbers. Even the power rankings are the model’s Elo, sorted — nobody here has a ballot. As the season plays, the 32 team trackers fill in automatically from the data.

The season-preview desk

• Weekly predictions, publicly graded — a probability for every game from a backtested, fully-stated model, frozen before kickoff; Week 1 is already locked.
• Power rankings, computed not voted — all 32 by the model's current Elo, movement tracked against frozen weekly snapshots, tier lines drawn by the gaps themselves.
• The Week 1 slate, game by game — every opening matchup's head-to-head ledger, the openers the schedule-makers circled, and what a Week 1 result is actually worth.
• The 2026 strength of schedule — and how much preseason SOS actually lies (the 32-team table, plus a 23-season back-test of the method itself).
• Reading the 2026 schedule: rest edges, short weeks, and who the calendar favors (69 rest-gap games, the bye map, and what history says an edge is worth).
• All 32 team pages — schedule, rest ledger, 2025 record, franchise baseline.

The division desk

Eight previews, one per division — the 2025 ledgers, each division's 2026 rivalry calendar, and the schedule facts that will decide tiebreakers, with the roster questions left honestly to September.

The evergreen reads that earn their keep in September

• Does Week 1 mean anything? — what 861 openers actually predict (less than the takes will claim).
• Rest & scheduling effects and QB continuity — the structural forces under the standings.
• Home-field advantage — smaller than remembered, and still real.
• Pythagorean wins — the October tool for telling records from teams.

How the season gets covered

The bundled game log refreshes from the public nflverse dataset; when it does, every team page's results column, the standings, and the prediction ledger update with it. New this season: a fully-stated, backtested Elo model publishes a win probability for every game — frozen before kickoff, graded after, with the misses kept on the page. What you still will not find here: locks, betting advice, or an untracked take that could be quietly disowned by October. A prediction that can't be audited is a take; ours are audited by construction, and the SOS back-test is the reminder of why the humility matters.