QB Composite Rating

Six numbers, one score. Expected points added per game — counting designed quarterback runs, not only dropbacks — carries 45 percent, and expected points added per dropback carries 20, because moving an offense is the job and the two say it at different grains. Completion percentage over expected, taken from Next Gen Stats, carries 15 and is the accuracy check that does not care how far the throws travelled. Success rate, the share of dropbacks that gained enough to stay on schedule, carries 10. Expected points per dropback in the second half within one score carries 5, and QB-fault sack rate carries 5 and enters negative, because taking the sack yourself is a failure rather than a misfortune. Each input becomes a z-score across the same population of qualifying passers before the weights apply, which is why a rating reads in standard deviations: plus one is a standard deviation better than an average qualifying starter, and zero is that starter.

A quarterback line is six arguments at once, and reading them one at a time is how a passer wins each of them and still loses the season. Jared Goff finished 2024 second at plus 1.670 on the strength of accuracy — plus 5.1 completion percentage over expected, third of 40 — while ranking outside the top five in expected points per dropback; Joe Burrow threw 723 dropbacks, more than anyone, and his per-snap efficiency has to be read against that volume rather than beside a passer with 400. Weighting the six is a way of holding them together, and the weights are stated so a reader can disagree with them precisely rather than in general. Sample size is corrected before the weighting: every rate is pulled toward the league mean by how many snaps it rests on, so a passer who has just cleared the qualifying floor is not ranked against a full-time starter on equal evidence.

The weights are Muffed's, not football's, and nothing outside this site uses them — a reader who thinks late-game play is worth more than five percent is not wrong, and the raw columns are published beside the score so that argument can be had with numbers. Three of the six inputs are expected points measured three different ways, so they move together and the score is less independent than six inputs sounds. And it describes a season that has happened. It is not a projection, it does not price a quarterback against anyone's expectations, and it says nothing about next week.