Fifteen games, one tie, and a scoreboard that in several places did not match the underlying football. Here is what week 5 of the 2026 season — the 65th through 79th games of a 272-game regular season — actually carried signal on, and what looked loud but wasn't.

A reminder before we start: a team runs roughly 55 to 85 offensive plays in a game. Every rate below is built on that, and none of it is opponent-adjusted. One week moves the needle; it does not reset it.

Table of Contents

1. The most informative result on the slate
2. What week 5 actually told us
3. What looked like news and wasn't
4. What to watch next week

The most informative result on the slate

Drake Maye, 35 dropbacks, +0.84 EPA per dropback, 374 yards, 4 TDs, 0 INTs, CPOE +22.6 in New England's 45–17 home win over Las Vegas. The Patriots entered the week as 3.5-point home underdogs and won by 28.

The reason this is the single most informative number on the board is not the margin — it is the gap from Maye's own four-week baseline. Through weeks 1–4 he had thrown 128 passes for 854 yards, 4 TDs, 7 INTs, with a CPOE of +2.3 and an EPA per dropback of –0.10. In other words, a replacement-level month followed by one of the best single-game passing lines of the season on a comparable sample of attempts (35 vs. a four-week average of 32).

One game is still one game. But the Patriots as a team posted +0.40 EPA per play on 66 snaps, the best offensive mark of the week, against a Las Vegas defense that had actually been stingy through the first month — it had held opponents to roughly –0.10 EPA per play across weeks 1–4, one of the tighter marks in the league, before New England hung +0.40 on it in one afternoon. If Maye's October 11 looks anything like the next four weeks, the four-week baseline was the noise and this is the correction. That is a real open question. It wasn't one before Sunday.

What week 5 actually told us

Pittsburgh's ground game carried a 12-point road win. Pittsburgh beat Indianapolis 34–22. The Steelers' rushing offense posted +0.32 EPA per carry on 22 carries, with a 55% success rate and two runs of 10-plus yards; their passing was merely good at +0.13 EPA on 32 dropbacks. Two of Pittsburgh's four touchdowns were Jaylen Warren plays — an 80-yard catch-and-run and a 50-yard run from scrimmage. The scoreboard (a 12-point road win) implies a bigger per-play edge than Pittsburgh actually had: the Steelers' net EPA margin was +0.21 per play, with Indianapolis's offense essentially neutral (+0.00 overall: –0.00 on 45 dropbacks, +0.02 on 24 runs). The signal: Pittsburgh's ground game produced two explosives in a 22-carry sample. The caveat: 22 carries is 22 carries.

Houston turned a field-position edge into a lopsided score. Houston beat Tennessee 26–3. Houston's offense was +0.09 EPA on 63 plays — good, but not blowout-good. Tennessee's offense was –0.39 EPA on 58 plays with two turnovers. A chunk of the difference lived in starting field position: on first-and-10 snaps, Houston averaged a start around its own 55-yard line; Tennessee averaged around its own 40 — about a 15-yard gap before either offense ran a play. Houston went 6-of-12 on third down; Tennessee 4-of-11. The real news here is a Titans offense that produced three points on nearly 60 plays, a datapoint against a Houston defense that had been middling across the first month.

Detroit outplayed a home favorite by 24 points. Detroit won 45–21 at Arizona, and the Lions were actually 5.5-point road underdogs on the closing line — not the favorite the final score might suggest. Detroit posted +0.33 EPA per play on 65 snaps, second-best offensive mark of the week. Arizona's offense, the team that was favored, managed only +0.02 on 71 plays — almost exactly replacement level. This is the cleanest result of the week: the team the market liked less was simply better, and the scoreboard caught up to the process.

New England's defense had a quietly strong week too. Las Vegas ran 57 plays at –0.02 EPA/play; their rushing was particularly poor at –0.34 EPA on 18 carries. New England has now shown an offense capable of +0.40 EPA/play in a game. That is a bigger shift from its four-week identity than any team on this slate.

Seattle's five turnovers explain the scoreline more than the three-point loss does. Seattle lost 32–29 at home to San Francisco. Seattle's offense came in at –0.43 EPA per play on 60 snaps, the worst mark among all 30 team-games on the slate, with two interceptions and three lost fumbles (five turnovers total). San Francisco's offense was itself only –0.11, so this was a game San Francisco won largely by being less bad. A three-point final score is not the whole picture here; a five-turnover home performance is.

The Giants and Commanders tied, and the efficiency numbers tied with them. Washington and New York played to a 27–27 tie. Washington posted +0.11 EPA on 61 plays, the Giants +0.06 on 84 plays. The Giants ran 23 more offensive plays than Washington and could not convert the volume into points. This is the rare game where margin and process agree exactly: there was no winner.

What looked like news and wasn't

Chicago's turnovers turned a competitive process into a blowout score. Green Bay beat Chicago 31–7 at home. Pull the per-play numbers and Green Bay was +0.20 EPA/play, which is good; Chicago was –0.34 EPA/play, which is bad. That part matches the scoreboard. But the Bears' pass offense produced –0.51 EPA on 44 dropbacks with 3 interceptions and 2 lost fumbles, and Chicago converted just 4 of 14 third downs. Five giveaways against a defense that didn't otherwise dominate is the story here — not a runaway Packers performance. Green Bay's own pass game (+0.50 EPA on 34 attempts) was legitimately excellent, but the 24-point margin leaned heavily on takeaways that are not a sticky skill from week to week.

Both Denver and the Chargers had rough offensive days — Denver's was just slightly less rough. Denver beat the Chargers 23–17. Denver's offense posted –0.27 EPA per play on 57 snaps; the Chargers' offense came in at –0.39 EPA on 63 snaps, with Justin Herbert sacked five times, going 14-of-28 for 121 yards with an interception at –0.57 EPA per dropback. Denver winning by six is not evidence Denver's offense functioned; it is evidence the Chargers' offense functioned less. The headline reads like a Broncos signature win. The underlying numbers read like a game where both offenses were among the worst of the week.

Pittsburgh's margin outran its efficiency numbers. The Steelers' net EPA in their 12-point win over Indianapolis was +0.21 per play — the eighth-largest of the week's 15 games, well behind Green Bay's +0.54, Houston's +0.48, New England's +0.42, San Francisco's +0.31 escape in Seattle, and Detroit's +0.31 upset in Arizona. A 12-point road win built on two long Jaylen Warren touchdowns is a real result; it is not a signal that Pittsburgh was the week's best team. Opponent adjustment matters too: this version of Indianapolis's offense was neutral, a step down from the team that scored 23 against Baltimore's 41 in Week 1.

What to watch next week

The question Maye's game opened — baseline or correction — is the one with the most riding on it. Thirty-five dropbacks at +0.84 EPA is not enough to overturn 128 dropbacks at –0.10, but it is enough to make week 6 a genuine test. If the next 30-plus dropbacks come in anywhere above neutral, the four-week baseline starts to look like a different quarterback than the one playing now.

The same is true in the opposite direction for the Chargers. Herbert being sacked five times at –0.57 EPA per dropback in a game his team lost is the kind of result that invites a one-week shrug. One more week of similar protection numbers against a different front would turn a one-game sample into a two-game pattern, and that is where early-season signal starts to firm up.

And in Chicago, the question is simpler and smaller: can an offense that generated –0.51 EPA on 44 pass attempts with five giveaways and a 4-of-14 third-down day do better without the turnovers? The underlying passing wasn't just unlucky — it was shaky on money downs too. One clean week would be evidence; another five-turnover afternoon would not be noise.

One week is one week. The reason to watch is that the sample doubles on Sunday.