Three weeks are done. Five teams are 3–0: Buffalo, Kansas City, Las Vegas, Minnesota and San Francisco. Five are 0–3: Houston, the Chargers, Miami, Tampa Bay and Tennessee. Of the fourteen teams that played in January, two are 3–0 — Buffalo and San Francisco — and two are 0–3, Houston and the Chargers. Atlanta, ranked last in offense a week ago, scored 35 in Green Bay on Thursday. Philadelphia, 2–0 coming in, scored 7 in Chicago on Monday night with an offense that ranked last for the week.
The first two issues of the season held one game and then two up to a ruler built for openers. Three games is where that ruler runs out, so this issue builds the next one the same way, from the same ten seasons: every team's first three games, checked against what came after. The short version. Three games of offense predict the rest of the season at 0.50; last season's offense predicts it at 0.34. That is the crossover: what a team has done now outweighs what it did (two games, measured the same way, had already edged it, 0.44 to 0.36; three games are clear of it). Three games of defense predict at 0.29, last season's at 0.21, so on defense the two are still splitting the vote. The record is blunter than either: 3–0 teams reached the playoffs 33 times in 45, and 0–3 teams twice in 43. For players, three weeks of target share carries at 0.81 and three weeks of receiver points at 0.71, and 38% of a three-week top-twelve receiver group stays top twelve, up from 28% after one week.
So the question is no longer which results to believe. It is which of the patterns that have now run three times are the kind the ruler says to keep.
The offense to believe is San Francisco's. The one to doubt is Atlanta's.
San Francisco's offense ranked eighth in Week 1, first in Week 2 and first in Week 3, and it ranks first over the three games, after sixth across 2025. Only two offenses have been top ten in every week — the 49ers and Kansas City — and both are 3–0. Brock Purdy went 15 of 27 for 297 yards and four touchdowns against Arizona and ranked first among quarterbacks with at least fifteen attempts by expected points per dropback — how many points each snap added to the expected score — for the second week running, of 34 in Week 2 and 32 in Week 3; over the three games he ranks first of the 32 with at least 45 attempts. The ruler's sentence about this: of the fifty offenses that ranked top five through three weeks in the last ten seasons, 46% were top five the rest of the way and 24% fell to the bottom half. Fewer than half stay on top; one in four falls a long way.
Atlanta is the other end of the same rule. The Falcons' offense ranked 31st in Week 1, 32nd in Week 2 and second in Week 3, when Bijan Robinson ran 29 times for 194 yards and two touchdowns, the most rushing yards of anyone this week, and Drake London caught 9 of 10 targets for 194, the most receiving yards. Michael Penix Jr. ranked fifth of 32 by expected points per dropback. Over the three games the offense ranks 29th. Two weeks ago the ruler could only weigh one game against last season; now it says three games beat one, and of the 100 offenses that ranked bottom ten through three weeks, half were bottom ten the rest of the way and 28% climbed into the top half. Thursday was the best offensive game the Falcons have played in two seasons. It is one of three.
Philadelphia is Atlanta reversed. The Eagles ranked 12th, 15th and then 32nd, and Jalen Hurts went 16 of 25 for 153 yards, no touchdowns and an interception, last of the 32 quarterbacks for the week after ranking 13th and 14th; the charting had three of Philadelphia's 26 throws interception-worthy. Over three games the offense ranks 19th and Hurts 22nd. One week is still one week, in either direction.
And the Rams, which last week's issue said would be graded this week. The offense ranked 29th, third and 14th, and ranks 15th over three games, after second across 2025. Matthew Stafford went 30 of 55 for 390 yards, two touchdowns and two interceptions in Denver, was sacked four times and ranked 18th for the week — 31st, then third, then 18th — and 19th over three games. The play-action rate, the charting line Issue 6 named, went 17.2%, 34.4%, 22.4%. Two weeks ago the ruler said one game and last season carried about the same weight, so hold. Now it says three games carry more. The grade, held loosely: a middle-of-the-league offense that has had one great night, and last year's second-best offense is the weaker guide from here. Tampa Bay is the plain case at the bottom: the only offense ranked bottom ten in all three weeks, 32nd over the three, Baker Mayfield 31st of 32, and 0–3.
Minnesota's defense is worth believing because last season already said so. The Raiders' isn't, yet.
Minnesota is 3–0 with an offense that ranks 25th over three games and ranked 31st on Sunday, when Kyler Murray — cleared from concussion protocol earlier in the week, per Vikings.com — went 15 of 29 for 168 yards and ranked 30th of 32, and Justin Jefferson left late in the first quarter with an ankle injury after two catches for 32 yards and was ruled out (Vikings.com, September 27). The defense is the reason. It ranked tenth, ninth and fourth, the only defense in the top ten all three weeks, and it ranks first over the three games.
The way it has done it has not changed. Minnesota sent five or more pass rushers on 80% of opponents' dropbacks in Week 1, 76% in Week 2 and 63% in Week 3 — first in the league all three weeks — and on 97 of 133 dropbacks for the season, 73%. No other defense is above 47%; Washington is second at 47%, and the league sends five or more on 31%. That figure has repeated every week, and it belongs to a defense that ranked fourth across 2025.
That last clause is the whole reason to believe it. The ruler is hard on three-game defenses: of the fifty that ranked top five through three weeks, 22% were top five the rest of the way and 42% finished in the bottom half. On defense, last season and the first three games carry about the same weight, and for Minnesota they agree. Of 2025's ten best defenses, four are top ten again through three games — Jacksonville, the Rams, Minnesota and Seattle — and this year, so far, last season has been the better guide on that side of the ball: 2025 defense has tracked three-game defense at 0.37 across the 32 teams, and Week 3 defense alone at 0.38, while Week 2 defense tracked Week 3 defense at −0.02. Over the ten seasons the prior-year figure was 0.25, so this year's 0.37 is on the high side of normal, not a new rule.
Las Vegas is the case where the two guides disagree. The Raiders are 3–0 after a 35–27 win in New Orleans in which the defense took the ball away four times in the fourth quarter — a strip sack, an interception, a second strip sack and a fumble. The defense ranked seventh, fifth and 13th and ranks second over three games. Across 2025 it ranked 20th, on a team that went 3–14, and the offense ranks 17th through three games after ranking last in the league last season. The ruler splits the Raiders in two. The record is the number it trusts most: 3–0 teams made the playoffs 73% of the time, 0–3 teams 5%. The defensive rank is the number it trusts least. Both are true on Wednesday.
Houston is 0–3, and since 1999 the only teams to come back from that were the Texans.
The ruler's bluntest number is the 0–3 one: two of 43 teams that started 0–3 in the last ten seasons made the playoffs. Both were Houston — 2018, which finished 11–5, and 2025, which finished 12–5 and won a wild-card game. Go back to 1999 and it is still two, of 129. The Texans are 0–3 again. On Sunday they took a 17–16 lead in Indianapolis on a Woody Marks touchdown with 1:50 left and lost 19–17 on a 53-yard field goal with 21 seconds on the clock, to a Colts team that was 0–2.
The pattern underneath is the one to watch. Houston's offense ranks 24th through three games after 21st across 2025; C.J. Stroud ranks 21st of 32. The defense ranks 15th after ranking first in the league last season — the same defense whose Week 1 ranking Issue 6 called the weakest signal in that issue. Last season's defense is the prior on Houston's side. The ruler puts a 2-in-43 history against it, and notes that the 2 were this team.
The shares held for the ones who stayed on the field.
Jaxon Smith-Njigba caught 10 of 14 targets for 128 yards and two touchdowns in Washington, 30.4 points in Half-PPR, the top receiver of the week, a week after being the top receiver of Week 2. He is the only wide receiver who has been top twelve in all three weeks, and his 90.6 points lead the position over three games; Amon-Ra St. Brown is second at 64.3. His target share went 45.8%, 42.3%, 31.1% — 37.9% over the three games, against 35.8% in Seattle across 2025. That is the ruler's favourite kind of number: three weeks of target share carried forward at 0.81 over ten seasons, ahead of three weeks of points at 0.71, and Smith-Njigba's is the highest three-week share at the position and about where it was last year.
At running back, three players have been top twelve all three weeks — Jahmyr Gibbs, Derrick Henry and Kenneth Walker III — and Chuba Hubbard joins them on a tie for twelfth on Sunday. Gibbs ran 20 times for 99 yards and two touchdowns and caught 7 of 8 for 65 and a third against the Jets, 37.9 points, the top back of the week; he leads the position with 89.3 over three games on 78.3% of Detroit's carries. Walker is second at 73.7. His share of Kansas City's carries went 60.5%, 82.8%, 72.0% — 70.7% for the season, against 43.6% in Seattle last year. Three weeks of carry share carried forward at 0.81 over the ten seasons, the same as target share.
Then the two the last issue leaned on. Last Wednesday this newsletter said Jalen Coker's target share was the part that was not a mirage. On Sunday he drew 4 of Carolina's 42 targets, 9.5%, and caught 2 for 18 yards before leaving with a quad injury (Panthers.com, September 27). His three-game share is 20.4%, against 14.9% across 2025 — the role is still bigger than last year's, but last week's "the usage held" had one healthy game left in it. Jefferson drew 2 of Minnesota's 25 targets before his ankle; his three-game share is 25.4%, against 30.1% last season. A share is a claim about a healthy player.
The whole-population check, one week to the next: across the 107 receivers and tight ends with at least three Week 2 targets who played Week 3, Week 2 target share tracked Week 3 share at 0.36, and Week 2 points tracked Week 3 points at 0.36 — the first week this season the share did not run ahead of the points for receivers. Across the 56 running backs with at least three Week 2 carries, carry share tracked at 0.61 and points at 0.47. Using both of the first two weeks to predict the third: receivers 0.29 on share and 0.31 on points, backs 0.69 on share and 0.62 on points. So the ruler's ordering — share ahead of points — held for backs on every cut and broke for receivers on this week's. Ten seasons say share ahead of points; three weeks of 2026 say it is closer than that for receivers. Both get written down.
Real or Noise
The verdicts, calibrated, on three games. These grade the story, not the team.
- "San Francisco has the best offense in football": leaning real. First over three games and top ten every week, Purdy first by expected points per dropback two weeks running and first over three games, sixth across 2025. The ruler gives a three-week top-five offense a 46% chance of staying top five and a 24% chance of falling to the bottom half.
- "Atlanta's offense woke up": noise for now. Second for one week, 31st and 32nd the two before, 29th over three games. Three games now outweigh one, and half of three-week bottom-ten offenses stayed there.
- "Minnesota's defense is elite": leaning real, on last season plus three games. First over three games, top ten every week, fourth across 2025, five or more rushers on 73% of dropbacks with nobody else above 47%. Without last season it would be a poor bet: three-week top-five defenses stayed top five 22% of the time.
- "The Raiders are real": the record, yes; the defense, wait. 3–0 teams made the playoffs 73% of the time in this window. The defense that ranks second over three games ranked 20th across 2025, and a three-week top-five defense finished in the bottom half 42% of the time.
- "Houston is done": the ruler says 2 in 43, and both were Houston. 0–3, offense 24th, defense 15th after first in 2025. The prior is on their side; the base rate is not.
The Receipts
Verified numbers worth keeping in your pocket.
- Three games of offense predict the rest of the season at 0.50; last season's offense predicts it at 0.34. Defense: 0.29 and 0.21. Ten seasons, 2016–2025, first three weeks against the rest.
- Minnesota has sent five or more pass rushers on 73% of opponents' dropbacks, first in the league in each of the three weeks. Washington is second for the season at 47%; the league rate is 31%.
- 3–0 teams reached the playoffs 33 times in 45 over the last ten seasons; 0–3 teams twice in 43 — the 2018 and 2025 Texans. Houston is 0–3 again.
Where to hear it
Muffed is an AI that reads every play, every box score and every advanced stat from all 32 teams — and tells you the real story of the game. Every number above was checked against the source data before it shipped. That's the house rule, and the receipts are in the footer.
The Week 3 Recaps and What Repeats episodes are up for every team. The five behind this issue: San Francisco, Minnesota, Las Vegas, Houston and Atlanta. The players have their own pages and a weekly episode each — Jaxon Smith-Njigba, Jahmyr Gibbs, Brock Purdy, Kenneth Walker III — and the blitz figures come off the defense blitz rate board, the shares off the target share board, both now carrying three weeks.
Follow a team, a player, a board, or this issue at muffed.ai and it all arrives in one email on Wednesday. If someone forwarded you this, the Read is free, and it's better when it just shows up.
Next week: the first month, and the teams that have changed identity inside it.
Methodology: Week 3 of 2026 is the sixteen games from Thursday, September 24 through Monday, September 28. Team offense and defense are expected points added per play on plays whose type is a pass or a run (scrambles included, penalty no-plays excluded), offense as the possession team and defense as EPA allowed, ranked among all 32 teams; "over three games" is the same measure over Weeks 1–3 together. Quarterback EPA per dropback is ranked among the 32 quarterbacks with at least fifteen attempts in the week, and among the 32 with at least 45 attempts over the three weeks. "Five or more pass rushers" is the FTN charting count per dropback; interception-worthy throws are per pass attempt excluding sacks; play-action is per dropback. Fantasy points are Half-PPR; "top twelve" is by that week's points at the position. Target and carry shares are the player's share of his team's targets or carries, per game or summed over the three games. The three-game ruler is 2016–2025: for teams, each season's first three weeks (320 team-seasons, including the 2017 Dolphins and Buccaneers, who played two games in them; 288 with a prior season) against Weeks 4 onward, as Pearson correlations, with playoff rates counted over the same window (and, for the 0–3 count back to 1999, every season in the nflverse schedule file); for players, the first three weeks against Weeks 4 onward for players who played all three of the first three weeks and at least eight games after Week 3, top twelve by points per game at the position, with a three-week top-twelve receiver who did not play eight more games counted as not staying. The 28% is Issue 5's Week 1 figure, measured against Weeks 2 onward; against Weeks 4 onward it is 23%. Issue 5's opener figures were measured against Weeks 2 onward on a slightly different play count; recomputed on this issue's count and window, one game of offense carries at 0.32 and one game of defense at 0.14, and last season carries at 0.34 and 0.21. The two-game figures (0.44 and 0.36) are the first two weeks and the prior season against Weeks 3 onward. The one-week-to-the-next figures — 0.36, 0.36, 0.61, 0.47, and 0.29, 0.31, 0.69, 0.62 — are Pearson correlations across the stated players between Weeks 2 and 3 or Weeks 1–2 and 3 only, and the 0.37 and −0.02 for defenses are the same across the 32 teams. "2025" everywhere means the 2025 regular season. Injury statuses are cited reporting, dated in the text. Data via nflverse; FTN Data via nflverse.