Methodology · NFL
How the NFL model works
Playoff seeding and Super Bowl odds for all 32 teams.
- Version
- v1 (market-anchored, + disagreement overlay)
- Updated
- Weekly in season
- Data source
- Betting-market win totals, spreads & totals; optional documented overlay
- Calibration
- Benchmarked against the market baseline; graded on the Ledger
The question
Each team's odds to make the playoffs, win its division, earn a bye, and win the Super Bowl.
The data
The betting market — season win totals plus weekly game spreads and totals. In the NFL the market is extremely efficient, so it's the honest starting point rather than a rating we invent from scratch.
How strength is measured
Team strength is implied from market prices. The board self-prices roster and situation because the market already has.
Recency & regression
Lines are re-pulled weekly; as the market moves, the ratings move with it. No manual decay curve — the market does the forgetting and remembering.
Injuries & roster
Mostly priced automatically — a key injury shows up as a line move the same day, so it flows in without a manual edit. For the rare case where we think the market is slow or wrong, a documented overlay lets us adjust a team's anchored win total with a written reason and date (a deliberate, scoreable disagreement with Vegas). Used sparingly, since the NFL market is the sharpest of the five.
Home advantage
Captured in the game model when turning a rating gap into a win probability.
The simulation
A Monte Carlo plays the season out thousands of times to produce the playoff field, seeding (seven teams per conference, one bye each), and a full bracket through the Super Bowl.
Postseason format
14-team playoff: seven seeds per conference, the No. 1 seed byes, then single-elimination through the conference championships and Super Bowl.
Backtesting
Beating the NFL market outright is famously hard; we track our numbers against the market baseline and grade them on the Ledger rather than claim an edge we haven't earned.
Known weaknesses
Because it's market-anchored, the model is roughly market-efficient — the value is in aggregation, seeding scenarios, and bracket paths, not in out-predicting Vegas on a single game.
Every published number is timestamped to the Ledger the moment it's made and graded against results — nothing here is backdated. Percentages are rounded for readability.