21+ only. Information and entertainment, not betting advice. Gambling problem? Call 1-800-NEXT-STEP (Arizona) or 1-800-GAMBLER.

AZMango

How it works

Five steps, all automatic, re-checked every week.

  1. Start from the market. Sportsbook lines are the best public forecast there is. We take the consensus spread, total and moneyline (with the sportsbooks' margin removed) from a daily snapshot, and more often on game days.
  2. Rate every team. Our power ratings combine season-long betting lines, recent scoring margins, offensive and defensive efficiency (EPA per play), injuries weighted by how much each player plays, quarterback changes, rest, travel and kickoff time. They work for any two teams, even when no line exists.
  3. Blend, carefully. Our ratings may nudge the market line by up to 3 points, but only when the weekly backtest shows those nudges helped. Right now they don't, so the guard rail has them switched off.
  4. Simulate. We play each game 10,000 times, drive by drive, so final scores land on 3 and 7 the way real NFL games do. That gives win chances, cover chances and a realistic range of scores.
  5. Calibrate and grade. A calibration step makes "70%" mean 70%. Every call is logged before kickoff and graded in public on the track record.

The reasons lists

Each "advantage" and "disadvantage" is a checkable fact: a league rank, an injury on the official report, the simulation, or the market. We list the strongest five each way. Small samples early in the season are noisy, and we say so.

What we don't claim

No model makes any game a sure thing, and ours does not beat the closing betting line. If you bet, use these numbers as one input, never as advice.

College football

The same five steps run for every FBS game, with three differences. There are 130+ teams that rarely play each other, so the ratings also use CollegeFootballData's Elo rating and yards per play, plus travel distance and time zones. There is no league-wide injury report in college, so we say so on every game page instead of guessing. And for any two teams that aren't scheduled, your browser works out the matchup from the same ratings and model weights. The college track record has its own backtest.

Current model version: 1. Data: nflverse (schedules, scores, team and player stats, official injury reports, snap counts), CollegeFootballData via cfbfastR-data (college schedules, scores, Elo, closing lines, play-by-play player stats) and The Odds API (odds snapshots).