Do NFL Domes Produce Higher-Scoring Games After Accounting for the Teams?

If you have been benching or starting skill players based on whether a game is indoors, the raw numbers back you up: dome and closed-roof NFL games through 2025 averaged 48.3 combined points versus 44.8 for outdoor or open-roof games. But once the study accounts for which home and away teams are actually playing, that 3.5-point edge flips to an adjusted difference of -2.1 points with a 95% interval from -5.7 to +1.5 and a Holm-adjusted p-value of 0.318. In other words, the dome scoring bump you see in box scores is largely a product of team composition, not the roof itself.

The sample covers 640 dome or closed-roof games and 1,487 outdoor or open-roof games from NFL regular seasons through 2025, all drawn from nflverse schedules and play-by-play data. The comparison controls for week, season, and the specific home and away teams in each game, so the adjusted estimates answer a sharper question: do indoor games still look different when the same teams are effectively compared against themselves across venue types?

For scoring, the answer is no. The unadjusted 3.5-point gap in combined points flips to an adjusted -2.1 points, and the 95% interval comfortably spans zero. The raw dome advantage is real in the data, but it does not survive once you stop letting dome-heavy offenses inflate the comparison. One caveat on uncertainty: the game-level bootstrap interval for combined points (+2.2 to +4.8 points) stays positive and excludes zero, while the team-adjusted HC3 interval spans it. The headline conclusion rests on the registered adjusted estimate, but the two uncertainty methods disagree on the raw comparison, which is exactly where team composition does the most work.

The passing-game outcomes tell the same story. Neutral early-down dropback rate, meaning how often teams pass on early downs when the game script is still balanced, averaged 53.9% indoors versus 53.7% outdoors, with an adjusted change of +1.9 percentage points and a 95% interval from -0.1 to +3.9 percentage points (Holm-adjusted p = 0.178). Deep-target rate, the share of passes thrown 20 or more yards downfield, averaged 10.9% versus 10.8%, with an adjusted change of +0.8 percentage points and an interval from -0.3 to +1.8 percentage points (Holm-adjusted p = 0.318). Neither clears the bar after multiplicity adjustment.

One outcome moved after adjustment, but in the wrong direction for dome optimists. Pass EPA per dropback, a per-play measure of expected points added on passing plays, averaged 0.038 indoors versus 0.037 outdoors, yet the adjusted difference was -0.050 with a 95% interval from -0.099 to -0.001 and a Holm-adjusted p-value of 0.178. The HC3 interval sits entirely below zero, but two pieces of evidence argue against reading this as a dome penalty: the Holm-adjusted p-value does not survive the family-wide correction, and the game-level bootstrap interval for this outcome (-0.019 to +0.022, bootstrap p = 0.92) straddles zero entirely. Treat this as noise, not a signal.

The practical takeaway for fantasy managers is restraint. The dome bump you notice in weekly box scores is mostly a who-is-playing effect, not a where-they-are-playing effect. These are retrospective associations from observed game conditions, not causal claims, and they say nothing about whether a pregame weather or roof forecast would improve your projections. Before auto-starting a receiver because his game is indoors, check who is under center and who is calling the defense; the roster is doing far more work than the roof.