The Cincinnati Bengals are 2-0. That part is easy to state and easy to verify — a 33-27 home win over Tampa Bay in week 1, then a 20-6 win at Houston in week 2, both confirmed in the schedule data with no ambiguity. What's harder is what to do with the way they got there — and specifically with the fact that the same defense that finished 2025 as one of the worst pass units in football is, through two games, allowing numbers that no full Bengals season has produced in the play-by-play era.

The sample is 127 total defensive plays (90 dropbacks, 37 rush attempts). Everything below has to be read through that lens, and it's the last time this piece will belabor it — but it needs saying once, clearly, up front.

What actually happened in weeks 1 and 2

Both opponents ran offenses that finished 2025 right around league average — Tampa Bay at -0.009 EPA/play, Houston at -0.031. Here's what the Bengals defense allowed:

Two things stand out. First, the pass rush produced four sacks in each game, against two different offensive lines. Second, the run defense number against Houston — -0.538 EPA per carry across 18 rushes — is the kind of figure that almost never shows up over a full game, let alone survives a season. Houston averaged 3.33 yards per carry, but the EPA is so much worse than the raw yardage suggests because Cincinnati was stuffing early-down runs against a team that was constantly playing from behind (Houston scored six points).

The comparison against 27 years of Bengals defenses

Here's why a 127-play sample is worth writing about: put it next to every full Bengals defensive season since 1999, and it currently sits on top.

Best defensive EPA/play in Bengals history (pbp era), including the 2026 start:

No full Bengals season in the pbp era has reached -0.222 over a thousand-plus plays. The best full year on record — the 2001 defense — is nearly a tenth of a point of EPA/play behind it.

Run defense is the more extreme piece of it. Best rush-defense EPA/play in Bengals history:

Every full Bengals season from 2018 through 2025 landed somewhere between -0.048 (2022) and +0.087 (2025) in rush EPA allowed — meaning the 2026 number, built on 37 carries, is operating in territory this defense hasn't touched in the last eight years, let alone matched over a full season.

Where 2025 actually stood

It's worth being precise about the other end of this comparison, because the framing "worst pass defense in modern Bengals history in 2025" overstates it slightly.

In 2025, Cincinnati allowed 0.165 EPA per dropback on defense — the third-worst mark in franchise history since 1999, behind 2019 (0.185) and 2002 (0.170), and essentially tied with 1999 (0.164). Across the league in 2025, the Bengals had the fifth-worst mark in pass EPA allowed (behind the Jets, Dallas, Washington, and Tennessee) and the fourth-worst mark in overall defensive EPA/play (behind Dallas, the Jets, and Washington).

So this isn't "historically worst defense → historically best start." It's "genuinely bad, top-five-worst-in-the-league defense in 2025 → two games in 2026 with numbers no full Bengals season has ever posted." The size of the swing is what makes it interesting; both ends need to be stated correctly.

The full year-by-year picture, 2018 through 2025:

Four straight years of positive defensive EPA/play — meaning opposing offenses gained value on the average snap against Cincinnati — bookended the Super Bowl-run defenses of 2021-22 on one side, and on the other, whatever the first two weeks of 2026 turn out to be.

What this is, and isn't, evidence of

The honest read: through 127 non-penalty defensive plays across two games, the Bengals have played defense at a level no full-season Cincinnati team has ever sustained, against two offenses that finished 2025 near league average. That's a real thing that happened. It coincided with two wins — Cincinnati beat Tampa Bay 33-27 and Houston 20-6 — but a small-sample defensive outlier doesn't, by itself, explain a football result; special teams, offensive execution, and plain game-flow (Houston trailing most of the second game, for instance) all mattered too. It is not yet proof the defensive unit has changed. Defensive EPA is one of the noisier team stats at this sample size, and a two-game window can flatter almost anyone; a four-sack, minus-0.5-EPA-per-carry outing is exactly the kind of thing that regresses hard once an offensive line or a quarterback shows up that isn't playing from a 14-point deficit.

What would make this a story rather than a data point is persistence. If the run defense is still south of -0.10 after week 6, that's a departure from anything the franchise has done in the pbp era. If the pass rush is still generating four sacks a week once the schedule stiffens, that's a real unit rather than two friendly matchups. Right now, this is two games where the numbers were extreme, against opponents that shouldn't have produced extreme numbers, from a defense that ranked among the league's worst four months ago. Worth flagging now, so the next month has a baseline to be measured against — and worth being skeptical of until it is.