What The Saturday Model is

What we build

• Our own preseason rating, fit on old seasons and tested on the seasons it never saw, labelled experimental because it is one.
• A pattern board: twenty-six published situational patterns, each with its full record, its sample size, and a plain paragraph on where the evidence is thin. The weak ones say so on the page.
• Team and coach pages for all 138 teams: schedules, rosters, portal moves, coaching histories, our number next to the market's.
• A proof page, graded every Sunday by an automated job, from board snapshots taken before kickoff. We grade what the site said before the game, never after.

What makes it different

Nobody beats the closing line, and we say so. Not us, not anyone selling you a lock. That sentence sits next to every performance number we publish.

Studies are registered before we look. Bands and kill conditions get written down first, so a lucky cell cannot dress up as a finding. Dead ideas go to a public graveyard and stay there.

Numbers must re-derive from code. Every published pattern record is the output of a script we can re-run. In July 2026 that standard caught five of our own published records that did not re-derive; four were corrected on the site the same day, one of them upward, and the fifth sits in a public rebuild queue until it reproduces. A site that corrects its own numbers in the open is the product.

The failures are free forever. The graded record, the graveyard, and every test we failed are never behind payment.

The 2026 season is the test

Everything above is history and method. The season that opens on August 29 is the forward test, graded in public, week by week. Backtests are the hypothesis; the season is the evidence.

A free account gets every rating, every team page, the graded record, and the digest on a 48-hour delay. Membership adds the live timing. Details and pricing live on the join page when subscriptions open.