Vance Joseph's defense has compressed the run game into a narrow, familiar band, allowing between −0.084 and −0.023 EPA per rush every single year — never league-worst, never league-best, always somewhere between "quietly stout" and "clearly above average." That baseline is why the last two weekends look strange.

Through 2 games of 2026, Denver has allowed +0.243 EPA per rush on 59 carries. That's not just a bad start against the ground — it's on the other side of zero from anything this coaching staff has fielded in four years.

The four-year baseline, and the break

The 2026 mark is roughly a quarter of an EPA per carry worse than the worst full season in the window. For context on scale: across the 16 games played league-wide in each of the first two weeks of 2026, defenses have allowed −0.048 EPA per rush on average. Denver sits about 0.29 EPA/rush above that league average — worst in the league through week 2. The next-worst defense against the run, Indianapolis, is at +0.221 on 63 carries, which is at least a reminder that early-season rush numbers can swing hard for more than one team.

There's a per-play success-rate echo, too: opposing rushers have posted a 44.1% EPA-positive rate against Denver's front, above the 37.4%–43.4% range in each of the last four years. That's directionally consistent with the EPA jump rather than a lone number carrying the split.

The pass defense is doing what it always does

The reason this is worth flagging — and not just filed as "small sample, bad opener" — is that the other half of the defense hasn't moved.

Denver's pass defense through two weeks is essentially indistinguishable from its 2024 and 2025 self. Whatever is off, it looks isolated to the run — a one-sided break rather than a general defensive collapse.

Two very different opponents, same shape

The tempting explanation is "Kansas City happened." That's partially true. Kansas City running back Kenneth Walker carried 23 times for 173 yards (7.5 YPC, a 60-yard long) in the week 1 loss at Arrowhead, and Patrick Mahomes added 4 rushes for 27 yards at +0.79 EPA per carry as a scrambler. Denver gave up 35 carries for 224 yards and +0.366 EPA per rush in that game, including five runs of 10-plus yards.

But week 2, at home in a 20–13 win over Jacksonville, is the part that keeps this from being a one-game footnote. The Jaguars ran 24 times for 95 yards (3.96 YPC) — the raw line looks like a normal defensive Sunday — but the EPA came in at +0.062 per rush, still worse than any of Denver's four full-season baselines. Bhayshul Tuten went for 65 yards on 13 carries, including a 16-yard run, one of two Jacksonville rushes that gained 10-plus yards on the day. Denver won the game; the run defense still lost the down-to-down math.

Two very different offenses moved the ball more efficiently on the ground than Denver's baselines allowed on average. That's the part of the small sample that carries the most weight: it isn't one back or one scheme.

How much of this is real yet?

Honestly, we don't know. Fifty-nine carries is a small sample, and EPA per rush is a noisy stat over that many plays; the gap between Denver's own 2024 and 2025 marks (−0.084 vs. −0.028) is a reminder that even full-season run-defense numbers move around inside the same coaching staff. The 11.9% explosive-rush rate is right in line with the range (10.1%–12.2%) and only a tick above 2025's 8.8% — so a piece of the EPA swing looks like early variance on which explosive runs happened to land, not a wholesale collapse of gap integrity.

The reason to flag it now instead of waiting: the shape is unusual. A defense whose pass numbers still look like the last two years, whose run numbers have inverted against two dissimilar offenses, and whose four-year baseline was as tight as Denver's, is worth checking on again in a few weeks. If the run EPA is still positive after four or five games — especially if the explosive-run rate climbs with it — the "Joseph run defense is reliable" prior that held will need to be rewritten. If it snaps back toward zero, this was a rough opener plus a workmanlike win, and the identity holds.

For now, the finding is narrower than a verdict: through 33 completed league games of 2026 and two Denver games, one half of the Broncos' defense is doing exactly what it has done for four years, and the other half is doing something it hasn't done in that span.