New to these reports? Start here
- Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
- "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
- Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
- A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
- If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.
Wind unders on FORECAST wind — pre-registration (2026-10-17, written before any forecast data was pulled)
Question: does "bet the under when wind is 15+ mph" survive when the wind is the FORECAST a bettor had before kickoff, rather than the observed game wind (biasbiasWhether the misses lean consistently one way. A projection can have a good average error size but still be biased if it is almost always too high. Bias is often the more fixable problem." data-def="Accidentally using information that was not available at the time. It makes predictions look brilliant and is the single most common way a backtest fools you.">look-aheadlook-ahead biasAccidentally using information that was not available at the time. It makes predictions look brilliant and is the single most common way a backtest fools you.) used in the heuristic scan?
Prior evidence (already seen, motivates the test, not part of it): - nflverse observed wind >= 15, outdoor: 2006-15 unders 56.8% (n=262), 2016-25 55.9% (n=204). - backtests/weather_ftn_value (2026-08, ERA5 reanalysis wind): holdoutholdoutData deliberately set aside and never looked at while developing an idea, then used once at the end as a fair test. Peeking at it first would defeat the purpose. wind >= 15 = 50.0% unders (n=44).
Data:
- Forecast: GFS MOS (MAV) station forecasts from the IEM archive (mos.py model=GFS), wind speed wsp
(knots, 10 m sustained) x 1.15078 -> mph, at the forecast hour nearest kickoff (3-hourly; nearest,
ties -> later).
- Run used: the latest MAV cycle whose run time is >= 6 h before kickoff (MAV is issued ~4 h after
the cycle, so it was public by ~2 h before kickoff). Variant B: latest cycle >= 24 h before kickoff.
- Station: nearest ASOS airport to each stadium (mapping in stations.py, fixed before results).
- Games: regular season + playoffs 2006-2025, roof outdoors or open, US stadiums only (no MOS abroad),
closing total and under odds from nflverse schedules (odds missing -> -110). Pushes = 0.
Primary rule (fixed): bet UNDER when forecast wind (primary run) >= 15 mph. Primary metrics: under win% and ROIreturn on investmentProfit as a percentage of the money wagered. +2% means $2 profit per $100 bet. at the recorded odds, 95% CI95% confidence intervalThe range the true value is plausibly in. If this range includes zero, we cannot rule out that the real effect is nothing at all.; report 2006-15 and 2016-25 separately. Pass = ROI > 0 with 95% CI excluding 0 over 2006-25 AND positive in both halves. Secondary (descriptive only, no pass/fail): thresholds 12 / 18 / 20 mph; variant B run; observed-wind rule on the same game set for comparison; correlationcorrelationHow closely two things move together, from -1 (opposite) through 0 (unrelated) to +1 (in lockstep). It does not by itself mean one causes the other. forecast vs observed wind.