Loading…
No games match this filter.
Final scores
Loading matchup…
Team ratings
| # |
|---|
Offense is points a team would score against an average defense; defense is points it would allow to an average offense, so lower is better. Overall is offense minus defense.
History
Every game the model projected, graded four ways.
Week by week
Records for each week with the filters above. Select a week to see just its games.
| Week | Games | ★ Our picks | Winners | Spread | Over/under | Moneyline |
|---|
Where the model is strongest
Spread record at or above each edge size, for the season and team selected. Small samples swing a lot; trust rows with a few hundred games.
| Edge | Games | Spread record | Win % | Over/under (same edge on total) |
|---|
How line movement played out
Movement is from the opening spread to the final one before kickoff. The covered columns show how often the team the line moved toward covered; the model columns show its spread record when it picked with or against that team.
| Line moved | Games | Covered opening line | Covered closing line | Model with the move | Model against |
|---|
How over/under movement played out
Movement is from the opening total to the final one. The covered columns show how often the direction the total moved (up means the Over) hit.
| Total moved | Games | Hit vs. opening total | Hit vs. closing total | Model with the move | Model against |
|---|
Sharp money and reverse line movement
| Signal | Games | Record | Win % |
|---|
Games
| Wk | Game | Final | Projected | Winner pick | Moneyline | Spread pick | Total pick |
|---|
How it works
Every team has an offense and a defense rating, measured in points and adjusted for who it played. Ratings start each season from a preseason estimate built from last season, returning production, roster talent, recruiting, the transfer portal and coaching changes, then move with every game.
A machine learning model adds what ratings miss: efficiency and style, rest, travel, altitude, time zones, the weather forecast at kickoff, and whether a starting quarterback is out. The projected margin is compared with the betting line, and the side the model likes is the pick. A strong pick is one where the model and the line disagree by at least 3 points.
How it has done in testing
Each past season was replayed week by week using only information available at the time.
| Season | Games | Against the spread | Winners | Over/under | Avg miss (pts) | Market miss |
|---|
Breaking even against the spread at standard odds takes 52.4%. Projections are for entertainment and analysis, not betting advice.