The Muffed Read — a weekly NFL newsletter written from verified data
Sports media gives you highlights and hot takes. Team analysts keep the real numbers in spreadsheets. The Muffed Read is the missing middle.
Every week, an AI reads every play, every box score, and every advanced stat from all 32 NFL teams — and writes you the real story of the week: who actually played well, how games were actually won, and which of the league's loudest narratives are real.
One rule makes this different from every other AI-written thing on the internet: every number is checked against the source data before it ships. The AI writes; it doesn't get to make up the stats.
Follow the Muffed Read. One story a week about what actually happened in the NFL — every number verified.
What's in an issue
- The lead
- One story: the week's best football, told as football, with the data as the evidence rather than the subject.
- Real or Noise
- Short verdicts on whatever the discourse is loudest about, in calibrated language — leaning real, leaning noise, never a certainty.
- The Receipts
- Two or three verified numbers nobody else is publishing, each with its rank and one line on why it matters.
No start/sit calls, no projections, no draft grades. Every figure is checked against the source data before an issue ships, and the methodology sits in the footer of each one — how we verify every number.
Every issue
How much should one game change your mind? San Francisco's four turnovers, Kansas City's 3 points, Tampa Bay's 43, Houston's overtime loss and Philadelphia's 2-5, run through a ruler built for Week 7.
Over ten seasons, a team's latest game has counted about 1.2 times as much as any earlier one, so after seven games it is roughly a sixth of the evidence. Good teams that lost in Week 7 won as often afterwards as good teams that won; middling teams that won by three touchdowns did win more.
Which records are running ahead of their points? Cleveland's and Denver's 4-2, Buffalo's 5-1, San Francisco's 6-0, Houston's 2-4 and Tampa Bay's six one-score games, run through a ruler built for six games.
Teams that had won a full game more than their points scored and allowed predicted kept about a tenth of that edge over the rest of the season; the gap carries at 0.12. But wins already won stay won: 6-0 teams reached the playoffs 7 times in 7 over the last ten seasons, and 4-2 teams that had been outscored 9 times in 13.
Which records are telling the truth? Minnesota's 5-0 with the 28th-ranked offense, San Francisco's win in Seattle, Houston's first win, Atlanta's fifth game and the ten 3-2 teams, run through a ruler built for five games.
At five games, a team's record, its point differential and how well it plays snap to snap predict the rest of its season about equally well. When they disagree, believe the play: 3-2 teams ranked top ten by net EPA per play reached the playoffs 25 times in 39 over the last ten seasons, and 3-2 teams in the bottom half 12 times in 33.
Which first-month signals are real? San Francisco's offense, Atlanta's run game, Minnesota's blitz, the five 0-4 teams and Jaxon Smith-Njigba's targets, run through a ruler built for four games.
Four games in, how a team plays repeats far better than how well it plays: blitz rate, run-pass lean and target share all carry into the rest of the season at 0.51 or better, defensive results at 0.29 and touchdowns at 0.19. And no 0-4 team in ten seasons has reached the playoffs; this season has five.
Which three-game patterns are worth believing? San Francisco's offense, Minnesota's blitz, the Raiders' 3–0, Houston's 0–3 and the receivers who kept their shares, run through a ruler built for three games.
Three games in, this season's offense now outweighs last season's. Defense still doesn't. Minnesota, the Raiders, Houston's 0–3, and the receivers whose shares held.
What repeated from Week 1 to Week 2? Chicago's 3, the Rams' 28, Carolina's defense and Jalen Coker's target share, run back through Issue 5's ruler.
Chicago scored 59, then 3. The Rams scored 7, then 28. Carolina allowed 59, then 3. The loud numbers flipped. The usage held — for Coker, for Walker, and for the ruler.
Which Week 1 results should you believe? Chicago's 59, the Rams' 7, Kansas City's defense and the Week 1 hero, run through Issue 5's ruler.
Seven of last year's fourteen playoff teams lost their opener. A 6–11 team beat a 14–3 team by three touchdowns. Two weeks ago this newsletter built a ruler for exactly this. Here's what it says.
How much does Week 1 actually tell you? Ten seasons of NFL openers, checked against what came next.
Both Super Bowl LX teams lost their opener. The biggest Week 1 winner missed the playoffs. And Week 1 winners still reach January at twice the rate of losers. All three are true.
Is "injury-prone" real? Five NFL seasons of who actually takes the field.
Christian McCaffrey's last five seasons: 7, 17, 16, 4, 17 games. Availability is the least repeatable thing we measure about a player — except in one group, where it runs an honest coin flip.
Which NFL teams got lucky in 2025? The luck audit, in 288 back-to-back seasons.
Kansas City won all eleven of its one-score games in 2024, then one all of 2025. Denver just went 11-2 in them. A decade of data on what a record proves — and what it only borrows.
How are NFL games actually won? The sport under the sport, in 19,811 charted dropbacks.
Pressure turned the league's best passer into a below-average one. Blitzing didn't buy what fans think. And the best pass in football came dressed as a run. The hidden layer of 2025.
What actually happened in the 2025 NFL season? The year, retold by the data.
Three teams went 14-3 and got there three completely different ways. The champion looked old-fashioned on purpose. And the league's most talked-about defense wasn't one. A look back at what really decided last season — before the next one starts.
Ask this data a question
Every figure on this page comes out of Muffed's verified panel — and you can query the whole thing in plain English. No account, no key. Ask the data →
Use Muffed in your AI
Ask Claude or ChatGPT football questions and get these verified figures back, with sources. Add https://muffed.ai/api/mcp — free, no account, no API key.