Passing and Scoring Evidence from Extreme Rain and Snow in NFL Games Through 2024

If you've ever benched a wide receiver because of a rainy forecast or chased a running back in a snow game, the honest answer through the 2024 season is that the NFL samples are still too small to tell you whether that move is justified. Four separate questions — heavy rain, rain plus wind, snow, and snow plus wind — each came up short of the evidence needed for a confident answer, so none of them earns a standalone verdict.

Every fantasy manager has a weather rule of thumb: rain kills the deep ball, snow turns games into run-fests, wind-plus-precipitation is a scoring pit. The problem isn't that these intuitions are wrong — it's that through the 2024 season, the league simply hasn't produced enough of these games to test them. We tracked four registered weather contrasts against preregistered sample minimums, and all four fell short. Heavy rain showed up in 39 qualifying games against a minimum of 75. Rain plus 15+ mph wind appeared in 19 games, 56 short. Snow showed up in 25 games, 50 short. Snow plus 15+ mph wind happened in just 3 games — 47 short of its 50-game floor.

That last number deserves a beat. Three games. Any pattern you spot in three games is a coin flip wearing a lab coat. Even the largest sample here — 39 heavy-rain games — is roughly half of what was preregistered as the minimum for a standalone article. These four conditions are kept strictly separate throughout: they are different weather events with different comparison groups, and their game counts are never added together.

What the descriptive numbers show — and what they can't

Because the samples are underpowered, every estimate below is descriptive context, not a confirmed effect. And critically, a wide interval is not evidence of no effect — it's evidence that we can't yet tell. Each estimate is adjusted for week, season, wind speed, and temperature, with its original 95% interval and Holm-adjusted p-value preserved.

In heavy rain, the deep-target rate was essentially unchanged (-0.0 points, 95% interval -1.1 to +1.0, Holm p = 1.000), deep completion rate was +1.5 points (-4.2 to +7.1, Holm p = 1.000), and EPA per deep target was -0.053 (-0.332 to +0.226, Holm p = 1.000). None of those intervals rules out a meaningful effect in either direction.

When rain and 15+ mph wind hit the same game, the descriptive estimates leaned toward less passing and less scoring: neutral early-down dropback rate -0.7 points (-5.5 to +4.0, Holm p = 1.000), pass EPA per dropback -0.046 (-0.143 to +0.052, Holm p = 1.000), field-goal accuracy -2.2 points (-15.8 to +11.4, Holm p = 1.000), and combined points -4.1 points (-9.6 to +1.4, Holm p = 0.585). The combined-points estimate is the most suggestive — a four-point drop would matter for game totals — but the interval still crosses zero comfortably.

Snow games showed a small descriptive shift toward rushing: neutral early-down dropback rate -0.8 points (-3.9 to +2.4, Holm p = 1.000) and rushing rate +0.8 points (-2.4 to +3.9, Holm p = 1.000) — mirror-image estimates that are exactly what you'd expect from a small, noisy sample. Pass EPA per dropback (-0.009, -0.099 to +0.082, Holm p = 1.000) and field-goal accuracy (-0.9 points, -10.5 to +8.8, Holm p = 1.000) were essentially flat.

The snow-plus-wind sample is the most extreme case of small-sample noise. Its neutral early-down dropback rate estimate is +5.0 points with an interval from +0.8 to +9.2 and a Holm p of 0.076 — on its face, a shift toward more passing in the worst conditions, which contradicts the usual intuition. But this rests on 3 games, the bootstrap interval includes zero, and the Holm-adjusted p-value does not clear a confirmatory bar. Combined points showed -6.6 points (-15.3 to +2.1, Holm p = 0.412), again on 3 games. Treat all of it as noise until more seasons accumulate.

Where the evidence stops

Two boundaries matter for how you use this. First, these are observed postgame conditions, not forecasts — a game that ended up in heavy rain tells us nothing about whether you should have seen it coming on Saturday. Second, these are associations, not causes: even with more games, this design describes what happened in bad-weather games, not what the weather did. And the four conditions are never pooled — heavy rain, rain plus wind, snow, and snow plus wind are separate questions with separate comparison groups.

The practical takeaway for fantasy managers: you don't yet have evidence strong enough to justify a weather-driven lineup change for these rare conditions on its own. If your league-mate benches a star receiver because of a snow forecast, the honest response is that nobody — including the data — knows enough yet to say who's right. As more seasons accumulate and these samples grow toward their preregistered minimums, the picture may sharpen. Until then, the smartest weather play is humility.