NFL moneyline and spread picks generated from a tier-sized quantitative model that projects game script, weighs quarterback matchups, weather, rest, and market mispricing.
NFL is a short, high-variance season — 17 regular-season games per team — which makes structural edges far more valuable than picking the right favorite each week. Our NFL model is built around that constraint: instead of producing a confident pick on every game, it grades each matchup and only releases a play when it finds a real edge — enough to beat the vig and absorb variance.
Each play comes with a unit size — 1u, 2u, or 3u — that reflects our confidence. Higher-confidence games get larger unit sizing; smaller edges stay small. No game gets bet at a flat unit "just because." The model fires, sized, or it stays out.
Football outcomes are dominated by a handful of structural levers that the betting market doesn't always price correctly. Every projection accounts for:
Most NFL games are decided by which team controls the script — leading early, running the clock, dictating pace. Our numbers reflect how each matchup is likely to play out, not just which team looks better on paper.
QB play is the single largest game-level lever in football. Every projection accounts for both starting quarterbacks and the pass rush they are facing. A backup or rookie debut is treated as a real change, not a footnote.
Wind and game-day precipitation can change a game, especially for passing offenses and totals. Every outdoor game is checked against the forecast before a play goes out. Dome and retractable-roof games get a clean reading; outdoor December games in Buffalo do not.
Thursday-night games on short rest, cross-country travel, and bye-week returns all shift true win probability, and our fair lines reflect them.
A play only goes out when our number and the market’s implied probability disagree by enough to beat the vig. The gap — not the raw "who wins" — is the entire game.
Plays are released once opening lines settle on Tuesday and again after Thursday's lineup confirmations. We do not publish blind picks before injury reports are official; availability swings a fair line by enough that guessing is worse than waiting.
Coverage runs the full regular season, including Thursday Night, Sunday slates, Sunday Night, and Monday Night. Slate volume varies — some weeks produce three to five NFL plays, some produce one. The model is calibrated to fire only on real edges, not to fill a weekly quota.