Every autumn, somebody tells you to bet the under because it is windy. The reasoning is sound — wind ruins the deep passing game and makes long field goals a coin flip — and the advice is almost always given without a number attached.
Here is the number. Across 5,024 outdoor regular-season games from 1999-2026, each additional mile per hour of wind is worth about 0.17 points off the total, or roughly 1.70 points per ten miles per hour. That carries a t of about -4.77, so the effect is real and not a story.
It is also small, and it is already in the price.
The effect
| Wind | Games | Unders | Overs | Under rate | Mean points vs total |
|---|---|---|---|---|---|
| 0-4 mph | 1,032 | 507 | 508 | 50.0% | +1.33 |
| 5-9 mph | 2,187 | 1,032 | 1,117 | 48.0% | +0.93 |
| 10-14 mph | 1,156 | 614 | 526 | 53.9% | +0.16 |
| 15-19 mph | 470 | 262 | 203 | 56.3% | -1.42 |
| 20+ mph | 179 | 99 | 78 | 55.9% | -0.99 |
Read the last column first. In the calmest games the total finishes +1.33 points above the line on average; above 15 miles per hour it finishes below it. That is the wind, and it moves in exactly the direction everyone says it does.
Now read the one before it. The under rate climbs from 50.0% to 56.3%, and a bet at -110 needs 52.4% to break even. So the highest bucket clears break-even, and this is the point in the article where most people would stop.
Why we are not claiming it
Pooling every game above 15 miles per hour: 649 games, with 361 unders, 281 overs and 7 pushes. That is 56.2%, a return of +7.35% per bet, and it sits 1.95 standard errors above break-even.
Under two. And that threshold was not chosen in advance — it is the edge of the bucket that looked best, picked after seeing all five of them. A result that only exists because of where the lines were drawn is not a result, and one bucket out of five at under two standard errors is what noise looks like when somebody goes hunting.
The honest summary is that wind lowers scoring by an amount the market has already accounted for, leaving a residual too small to bet with any confidence.
The check on our own arithmetic
Games played under a closed roof have no wind at all, which makes them a control. If this method were quietly picking up something other than wind — a late-season effect, a scheduling artefact, a change in how totals are set — roofed games would show it too.
1,757 roofed games finish +1.24 points against their total — which is the calm-weather figure of +1.33, not the windy one. Indoors behaves like a still day, because indoors is a still day. The coefficient is measuring wind.
What it is actually good for
Not a bet. The useful version of this finding is descriptive: a 15-plus mile-per-hour forecast is worth about 1.70 points per ten miles per hour of scoring, the books know it, and a preview that says "the wind is why this total is low" is telling the reader something true about a number they can see. That is worth writing. "Hammer the under" is not.
Across the whole sample wind explains 0.45% of the variation in how a total finishes. The other 99.55% is football.
Method
Every regular-season game in nerddawg.nfl_game_history with a final score, a closing total and a recorded wind speed, played outdoors or under an open roof: 5,024 games, 1999-2026, mean wind 8.44 miles per hour and a maximum of 71. A game with no recorded roof is counted as outdoors. The slope and correlation are ordinary least squares of the total's residual on wind speed. Pushes are excluded from under rates rather than counted as half, because a returned stake is not a half-won bet. The standard error for the pooled test uses the break-even rate, which is the conservative choice. Measured 2026-10-13 and recomputed whenever this page is rebuilt.
