How Temperature Changes the Way Teams Play
The last post found that temperature moves the total score. Warmer games score more, colder games score less. But that's the view from the scoreboard. This post goes under the hood: turnovers, passing yards, rushing yards, third downs and field goals. For each one we checked two things: does it change steadily as it gets hotter, and does it dip (or spike) at both extremes, shaped like a U? Same treatment rain, snow and wind got in the original research.
Heads up
We rewrote a lot of this post after fixing a weather-timing bug (story in this post). Two conclusions actually changed, not just numbers: passing's "best temperature" isn't a realistic number anymore, and the effect on who wins went from "nothing" to "something real."
The raw pattern
Passing climbs pretty steadily as it gets warmer. Rushing does close to the opposite, then bumps back up when it's really hot. Turnovers, third downs and field goals don't show much of a pattern at first glance. But raw averages don't account for team strength or home field, so every one of these has to pass a fair test before we believe it.
The pattern that vanished: field goals
The raw numbers looked like a sure thing: kickers made 66.2% of their tries below 20°F, versus 72–75% everywhere else. Cold air, frozen fingers, rock-hard ball. Makes total sense, right? Then we accounted for how good the teams and their opponents were, and the gap disappeared completely. It wasn't the cold making kickers miss. It's that weaker teams (and teams facing tougher defenses) happen to play a bigger share of their games in the cold. Turnovers and third downs told the same story: no real effect once the comparison is fair.
This is the lesson this site keeps learning: a pattern that looks clean in raw numbers isn't a finding until it survives a fair test. This one didn't.
What survived: passing and rushing (with one catch)
Rushing yards bottom out at 76.4°F and climb at both extremes, a real U-shape. That makes sense, and the low point sits right in the middle of normal football weather. Passing is where the bug fix changed things. Its curve is still real (an upside-down U, not a straight line), but its peak now lands at 104°F, which is hotter than almost any game ever played. So the honest takeaway for passing isn't "it's best in nice weather and drops in the heat." It's closer to passing just keeps climbing across the whole normal temperature range. The curve eventually turning down is a quirk of the math, not something you'd actually see on a Saturday.
The rushing story still holds up as first told: offenses lean on the run whenever throwing gets harder, whether it's frozen fingers in the cold or a slick ball and tired arms in serious heat.
Does any of this change who wins? This time, yes. A little.
This is the one spot where fixing the bug flipped a conclusion. Rain, snow and wind don't change who wins, and our first look at temperature said the same. The corrected data disagrees: both the straight-line and curved versions of the effect come back real (p = 0.023 and p = 0.016; a "p-value" under 0.05 is the usual bar for "probably not luck"). But the shape isn't a neat U. The raw pattern is flat through the cold and mild range, then climbs only when it gets hot:
Team strength still decides most games no matter what the thermometer says. This is a real but small effect sitting on top of that, and it's all about the heat, not both extremes.
See the numbers behind these charts on the dashboard, or read about an even bigger temperature swing, the gap between a game's temperature and what the visiting team is used to, in the next post.