How the ratings work
A rating compares teams by the points they have scored and allowed, adjusted for the opponents they have actually faced. It is a relative strength number. It is not a league standing, a point spread, or a guaranteed result.
Rating = capped average scoring differential + average rating of opponents played
What counts
Only final games enter the rating. A game that has not been played stays out. Each completed game counts once, and every game counts equally. The current NFL model does not add weight for how recent a game was, for home field, or for a team’s reputation before the season.
Capped scoring margin
For each team in each completed game, the margin is points scored minus points allowed, limited to a range of −21 to +21. A win by more than 21 counts as +21. A loss by more than 21 counts as −21. A tie counts as zero and still counts as a game played. The capped average is that team’s margins added together and divided by games played.
Strength of schedule
Strength of schedule is the average rating of the opponents a team has played, counting each game separately. Playing the same opponent twice counts that opponent’s rating twice. Because every team’s rating depends on its opponents, the ratings are solved together. They are then shifted so the league average is zero.
A positive rating is above an average team in this model. A negative rating is below an average team. The units follow the capped point margin.
Rank change
The rank-change column on the weekly chart compares the new order with the prior published snapshot. The chart itself names that comparison.
What the number leaves out
Injuries, lineup changes, weather, travel, and home field are not separate inputs. A short sample or an unusual schedule can move a rating. The difference between two ratings is not a win probability and not a predicted final score.