About
The 53 Report
We grade every NFL general manager. Same dataset, same definitions, every team since 2012.
32
teams
1.3M
data rows
1980+
draft picks
2012+
snap counts
3
article shapes
Open
source
Why this exists
Drafting and keeping are different skills.
Every NFL team has 53 active players. The decisions about who fills those slots, who stays on a second contract, and which draft picks become starters are the most important and least examined work in football. The general manager owns those decisions. Almost nobody grades that work with the same definitions across all 32 teams.
Most teams that draft well don't retain at the same rate. Some teams retain everyone and draft poorly. The gap is real, it's measurable, and the numbers tell a different story than the narrative. Our product is the grade that comes out the other side.
What feeds a grade
The four columns of every GM scorecard.
01
Hit rate
Share of the GM's picks that produced 500+ snaps in any single regular season, against the league baseline for the same window.
02
Retention
Share of those picks still on the current 53-man roster, both raw and snap-weighted.
03
Position construction
How the homegrown players distribute across position groups, read against the team's identity.
04
The franchise pick
The single bet that defines the regime. Did it hit? What does the rest look like if it didn't?
Article shapes
Three shapes, one dataset.
The full grade
A tenured GM graded across the four columns, with a final letter.
Grading Brandon Beane → NarrativeA paradox or anomaly
Worth its own piece, with named players and a counterpoint, but no final letter.
Roseman's Eagles record → MethodologyLeague-wide framing
A methodological cut across all 32 teams. No team focus.
Two teams clear both bars →How we work
Open data, no sponsors.
All data is public. Our methodology is documented at /methodology, the dataset is open-source, and every published article links to the query that produced it.
We don't run advertising. We don't take sponsored content. The newsletter is free.
How we built it
SQLite, normalized joins, public sources.
Pro-Football-Reference and the nflverse open-source community publish weekly rosters, snap counts, and draft records back to 1980. We pulled those into a single SQLite database, normalized the joins, and aligned the time series. Schema and ETL at github.com/pete-builds.
Every article goes through a stat-fidelity check before it ships. If a number on this site is wrong, it's a methodology question, not a "trust us."