Through Monday night of Week 5, the 2026 regular season is 79 of 272 games old — 29% of the way home. Every team has played four or five games. The average offense has run 303 plays so far (minimum 249, Kansas City; maximum 360, Chicago). For reference, a full 2025 season ran from 941 to 1,268 offensive plays per team. The sample we're working from is roughly a quarter of a season.

That's the whole frame for what follows. Five weeks is enough to notice things; it is not usually enough to overrule a full prior season, especially before adjusting for opponent. This is a survey of what the pooled 5-week data does and does not license us to believe — not a prediction of where anyone finishes. And because EPA/play is the one stat doing almost all the work in surveys like this, we also check a second, independent metric — success rate allowed — against the defensive storylines below, to see whether it tells the same story or a different one.

Table of Contents

1. What we think we know
2. Confirming it a different way: success rate allowed
3. Two storylines, judged on stability
4. Believe it or not yet

What we think we know

Six takeaways, each from a different team or unit, each with its own cumulative sample.

1. Minnesota's defense has been the league's stingiest through five weeks — and the slate was not elite. The Vikings' defense has allowed −0.205 EPA/play across 279 plays, best in the league through Week 5. That's a real move: in 2025 they finished at −0.101 EPA/play over 1,016 plays (3rd-best), so they improved from an already-strong baseline. The weekly line is steady — −0.17, −0.13, −0.25, −0.17, −0.35 EPA/play allowed every week. No single blowup is doing the work. Opponent-adjustment caveat: Minnesota's five opponents (GB, CHI, TB, MIA, NO) averaged −0.046 EPA/play on offense over 1,558 plays this year — a below-average slate. The number is real; part of the ranking is earned by schedule.

2. The Chargers' offense is the league's worst — and this one is stable. Los Angeles has posted −0.207 offensive EPA/play on 307 plays, dead last. In 2025 they were −0.035 over 1,139 plays (25th). That's a real collapse against a near-full prior-season sample, and it hasn't been one bad afternoon: weekly EPA has run −0.24, −0.26, −0.07, −0.09, −0.40 — five negative weeks in a row at a sample large enough (~60 plays per game) that each week is meaningful on its own. The opposing defenses were about league-average: LAC's five defensive opponents (ARI, LV, BUF, SEA, DEN) combined for +0.023 EPA/play allowed over 1,573 plays. This is not a slate story.

3. Buffalo and San Francisco are the two best offenses in football, essentially tied. Buffalo sits at +0.211 EPA/play (293 plays, 1st in the league), San Francisco at +0.209 (277 plays, 2nd). Both are up on 2025, when Buffalo finished 2nd-best at +0.138 (behind the Rams' +0.145) and San Francisco finished 9th at +0.061. Buffalo's weekly line (+0.23, +0.43, −0.02, +0.10, +0.29) is a continuation; San Francisco's (+0.18, +0.46, +0.49, +0.12, −0.12) is a bigger jump but the floor has lifted. One asterisk on Buffalo: its defense has been shaky over the same five weeks, allowing +0.134 EPA/play (350 plays, 30th of 32) — this is an elite offense carrying a defense that hasn't caught up yet, not a two-way powerhouse.

4. Dallas has the worst defense in football, again. Dallas's defense is allowing +0.193 EPA/play across 303 plays — worst in the league. In 2025 they were +0.167 across 1,027 plays, also worst in the league. The prior sample is more than three times the current one and tells the same story. This is a trend, not a sample fluke.

5. The Lions' defense is the biggest negative mover in the league. Detroit's defense has gone from −0.011 EPA/play (13th in 2025, over 1,039 plays) to +0.152 EPA/play allowed (355 plays, 2nd-worst) through five weeks — a +0.163 EPA/play swing in the wrong direction. The weekly line is loud: −0.03, +0.43, +0.09, +0.29, +0.02. Three of five weeks were bad. Detroit's five opponents (NO, BUF, NYJ, CAR, ARI) averaged +0.040 EPA/play on offense over 1,511 plays — but that's almost entirely Buffalo's elite offense pulling the average up; strip Buffalo out and the other four opponents (NO, NYJ, CAR, ARI) averaged essentially zero (−0.0004 EPA/play over 1,218 plays), dead league-average. The decline isn't an artifact of facing great offenses.

6. The Raiders' offense is the biggest positive mover. Las Vegas has climbed from −0.215 EPA/play (dead last in 2025, 941 plays) to +0.004 EPA/play (317 plays, roughly middle of the league) — a +0.219 EPA/play swing, the largest offensive improvement in the league off the smallest 2025 baseline. The weekly line (−0.04, −0.10, +0.11, +0.05, −0.02) looks less dramatic: closer to steady average play than to a breakout. The honest read is that the offense has climbed out of a hole, not that it's good — "mediocre" is the news here. But five weeks of near-zero after a full season of deeply negative is a real level change.

Confirming it a different way: success rate allowed

EPA/play rewards big, game-swinging plays, which can make a five-week sample noisier than it looks. Success rate — the share of plays that beat down-and-distance expectation — is stickier and less home-run-dependent, so it's worth checking the four defensive storylines above against it. League-average success rate allowed was 44.3% in 2025 and 43.7% through five weeks of 2026.

• Minnesota: 41.7% allowed in 2025 (already better than average) → 32.0% through 5 weeks (294 plays) — a huge additional drop. Two different metrics, same conclusion: this defense has taken a real step forward.
• Dallas: 48.7% allowed in 2025 (already worse than average) → 50.9% (324 plays) — still bad, slightly worse. Confirms "worst defense, again."
• Detroit: 42.5% allowed in 2025 (better than average) → 48.2% (371 plays) — now clearly worse than the 43.7% league mark. Independent confirmation of the collapse.
• San Francisco: 48.2% allowed in 2025 (worse than average) → 41.8% (335 plays) — now better than the 43.7% league mark. This matters because it's a full-sample number, not a single-week one — so whatever the EPA line owes to one big game, the success-rate shift says the day-to-day stop rate has genuinely improved too.

All four EPA stories hold up under a second metric. That's a modestly higher bar than most five-week survey claims clear.

Two storylines, judged on stability

The Chargers' and Eagles' offenses both fell off a cliff — on different timelines

Of the big movers, LA's offensive drop is the one most insulated from the usual early-season complaints. Every one of its five weekly EPA figures is negative, the sample is already 307 plays, and the opponent slate on defense (+0.023 EPA/play allowed, above) is close to average. In 2025 the Chargers were middling (25th on offense); through five weeks of 2026 they are last by a wide margin, with Tennessee (−0.149) and Philadelphia (−0.127) the only teams in the same neighborhood.

Philadelphia's offense fell differently: +0.021 EPA/play in 2025 (15th, 1,067 plays) → −0.127 through 5 weeks of 2026 (287 plays, 30th). The weekly line is +0.07, +0.04, −0.30, −0.29, −0.20 — a step-down after Week 2 that has now held for three straight weeks, not five. The opponent slate (WAS, TEN, CHI, LA, JAX) averaged essentially +0.002 EPA/play allowed over 1,467 plays — dead league-average defense. The drop is not opponent-driven, but it's a shorter streak than the Chargers' and worth re-checking rather than banking.

Philadelphia's defense moved the wrong way too: −0.070 EPA/play in 2025 (9th, 1,121 plays) → +0.069 through 5 weeks (330 plays, 24th) — a +0.139 swing. Success rate allowed tells a milder version of the same story: 42.6% in 2025 (better than average) → 44.5% now (348 plays), just above the 43.7% league mark — worse, but not the outlier the EPA number alone suggests. Both sides of the Eagles sliding at once in the same five weeks is unusual; it's one reason to treat this as two distinct, still-forming stories rather than one verdict.

San Francisco's defense has flipped — and it's not just one week

In 2025 the 49ers' defense allowed +0.070 EPA/play (23rd, 1,142 plays). Through five weeks of 2026 they're at −0.067 (317 plays, 4th-best) — one of the three largest defensive EPA improvements in the league. The weekly line matters, though: −0.26, 0.00, +0.17, +0.09, −0.43. Four of the first five weeks were middling to bad; Week 5 (vs. Seattle, 60 plays at −0.43 EPA/play allowed) is doing roughly half the season's aggregate work on its own. That's the part of the story that looks fragile. But the success-rate number — 48.2% allowed in 2025 down to 41.8% through five weeks, both full-sample figures — says the week-to-week stop rate has also genuinely improved, not just the EPA total. Real, but still more one-week-dependent than Minnesota's version of the same claim.

Believe it or not yet

Six reads, graded on how much the 2026 sample should actually move you.

• Vikings defense is a top-5 unit: believe it. The 2025 baseline was already strong (3rd), the weekly EPA line is uniformly good across 279 plays, success rate allowed confirms it independently, and even the opponent-adjustment haircut leaves them well above average. What would kill the read: back-to-back weeks allowing +0.10 EPA/play or worse against competent offenses.
• Chargers offense is one of the league's worst: believe it. Five negative weeks, a middling 2025 baseline (25th), opponents on defense roughly league-average. What would confirm it further: a sixth and seventh negative week keeping the season figure under −0.15.
• 49ers defense is now elite: lean believe, not fully there. One game is doing a lot of the EPA lifting, but the success-rate improvement (48.2% → 41.8% allowed) spans the whole sample, not just Week 5. What would confirm: two more weeks under −0.05 EPA/play allowed against non-trivial offenses. What would kill it: reversion toward the 2025 figure (+0.070) with Week 5 looking like the outlier.
• Lions defense has collapsed: lean believe, with reservations. The move is +0.163 EPA/play across 355 plays against a full 2025 season, with three of five weeks visibly bad and success rate allowed moving from better-than-average to worse-than-average. Stripping Buffalo from the opponent slate still leaves a real decline against average offenses. What would confirm: another two weeks above +0.10 against non-Buffalo offenses.
• Raiders offense is now competent: lean believe. The 2025 figure (−0.215) was so bad that the jump to zero is a level change the sample can support, and no single week carries it. "Competent," not "good." What would push it to good: three more positive weeks.
• Eagles offense has broken: lean believe, shorter leash than the Chargers. The three-week step-down (Weeks 3–5) is more concerning than the headline number, and the slate was not the cause (opponents allowed dead-average EPA/play). What would kill the read: a bounce-back pair against similar defenses. What would confirm: a fourth consecutive negative week.

A reminder on the frame. None of these 2026 numbers are fully opponent-adjusted — the opponent-slate figures above are the simple version of that correction, and even those small shifts (Minnesota's slate at −0.046, Philadelphia's at +0.002) matter at this sample size. A single additional week can move a team two or three spots either direction on a roughly 300-play base. Treat ranks as current standing, not as how the season ends. The teams on the "lean believe" side of the ledger above are the ones to re-check in two weeks.