NFL Wind Unders Explained

Football is a passing league, and passing is the part of the game wind disrupts most. A quarterback throwing into a 15-mph crosswind doesn't complete the same pass he completes in calm air: the ball drifts, deep shots get harder to time, and coaches lean on the run more than they'd like. All of that pushes scoring down. The question isn't whether wind matters. It's whether the number on the board already accounts for it.

What the data shows

Looking at outdoor-stadium NFL games from 2003 to 2025, games with observed wind of 10 mph or higher at kickoff went Under the total at a materially higher rate than games with less wind. The effect held up, monotonically, as wind speed increased. Games under 10 mph landed close to a coin flip against the total, exactly what you'd expect if the market were pricing calm-weather games correctly. Games at 10+ mph did not land at a coin flip. The gap between the two groups is the edge.

That gap didn't fade in more recent seasons either. If anything, the 2021-2025 subset of the data shows a larger edge than the full 2003-2025 sample, which argues against the theory that books have already caught up and priced this in.

Why the market misses it

Sportsbooks build a total around team strength, pace, and matchup: the things that stay true all week. Weather forecasts, by contrast, are only reliable in the final day or two before kickoff, and they change. A book that shaded every total toward the Under based on a 5-day forecast would get burned constantly when that forecast doesn't verify. The rational response for a book is to wait, which means the total often doesn't fully reflect the wind until very late, sometimes not until after significant money has already moved the number for other reasons.

That lag between "the forecast is confidently bad" and "the total has fully adjusted for it" is where this edge lives. It's a market-timing story, not a secret formula: we're not out-projecting the book on talent or scheme, we're acting on public weather data on a timeline the total hasn't caught up to yet.

The exact rule we use

We bet the Under on an outdoor-stadium game's total when the National Weather Service forecast shows sustained wind at or above our threshold at the kickoff hour, captured within 48 hours of kickoff. Two details matter, and we're not going to round them off:

• We buffer above the researched cutoff. The backtest above was graded against wind that actually showed up on game day: the observed wind. Our rule has to act on a forecast, which is not the same thing; forecasts miss. Rather than fire on the exact 10 mph line the backtest validated, our live rule requires 12 mph or more in the forecast. That buffer costs a small amount of the researched edge in exchange for real protection against forecast error. The trade-off was itself tested: the 12 mph cutoff still shows a clear, positive edge in the same backtest, just a slightly smaller one than the 10 mph line.
• We only act inside a 48-hour window. NWS hourly forecasts get materially less reliable further out. A game outside that window is simply skipped for now, not marked a permanent no-play. A later refresh, once the game enters the window, evaluates it fresh.

Indoor and retractable-roof stadiums are excluded entirely; wind isn't a factor if there isn't any.

What this is and isn't

This is one of three independent NFL signals we publish. The others are a separate totals projection model (versus the opening total, all season) and an early-season-only spread angle (weeks 1-4, versus the opening spread). Wind unders are the most mechanism-driven of the three: there's a physical, explainable reason wind suppresses scoring, which is different from a model finding a statistical pattern with a less obvious cause. When the wind rule and the totals model agree, we publish one Under pick. When they disagree, we publish nothing on that game's total rather than picking a side.

This is not a guarantee of anything. A forecast can be wrong, wind can die down by kickoff, and any single game can go either way regardless of the signal. The point of publishing the exact rule, in public, before the season starts, is that you can hold us to it. You don't have to take our word for how it performs.

See the record for yourself

Every NFL pick, wind under, totals model, or early-season spread, is graded publicly on the scoreboard, win or lose, with the pick and the reasoning both on the record before the game starts.