Research
Extreme heat and total scoring in non-conference college football
A widely repeated claim says non-conference FBS games go Over the closing total roughly 59% of the time when kickoff temperature is between 85 and 106°F. We tested it the boring way: every completed, outdoor, regular-season non-conference game we could source with a kickoff-hour temperature and a closing total, graded against that total, split into temperature bands, with the cooler-weather control group reported alongside rather than buried.
The headline
The 59% claim does not hold in our data. Non-conference games at 85–106°F went Over 52.8% of the time, against 49.1% for the sub-85°F control — no meaningful separation.
85–106°F Over rate
52.8%
132–118–6 O–U–P
250 graded games (minimum 40)
Control: under 85°F
49.1%
575–595–15 O–U–P
1170 graded games in the same seasons
Hot minus control
+3.7 pts
difference in Over rate
1,441 graded non-conference games, 2016–2025
Over/Under by kickoff temperature
| Temperature band | Games | O–U–P | Over rate | Avg points | Avg closing total |
|---|---|---|---|---|---|
| Hot: 85–95°F | 229 | 119–104–6 | 53.4% | 59.2 | 58.1 |
| Extreme: 96–106°F | 27 | 13–14–0 | 48.1%insufficient data | 57.6 | 58.7 |
| Control: under 85°F | 1185 | 575–595–15 | 49.1% | 56.0 | 56.2 |
| Cool control: under 70°F | 437 | 214–218–5 | 49.5% | 54.8 | 55.4 |
Over rate is overs ÷ (overs + unders), derived from the record shown — pushes are excluded from the rate but counted in the games column. Bands with fewer than 40 graded games are greyed out and never feed a headline. The cool-control row is a robustness cut and overlaps the sub-85°F control.
What the numbers say
The popular figure does not reproduce here. Non-conference games at 85–106°F went Over 52.8% of the time across 250 graded games, against 49.1% for the 1170 control games below 85°F — a difference of 3.7 points, well inside the noise you would expect on samples this size. Average scoring in the hot band (59.0) sits close to the average closing total (58.2), which is what a market that already prices heat looks like.
Splitting the range matters. If a single narrow slice carried the whole effect, the 85–95°F and 96–106°F rows above would disagree sharply — that is precisely the trap that produced a wrong headline in our first CFB wind write-up, and why every band here is shown with its own record and sample size rather than folded into one number.
Method
- Completed FBS regular-season non-conference games, 2016–2025, taken from the ESPN scoreboard. The conference/non-conference split uses ESPN's own per-game flag.
- Indoor and roof-closed venues are excluded outright.
- Kickoff-hour temperature comes from the Open-Meteo ERA5 reanalysis archive at each venue's coordinates, matched on the UTC kickoff hour.
- Totals are ESPN consensus closing numbers. Games without a closing total or a kickoff temperature are not graded.
- Bands: 85–95°F, 96–106°F, control below 85°F, plus a secondary sub-70°F cut. Readings above 106°F fall outside the tested claim and are excluded.
- Minimum sample of 40 graded games per band. Below that, the band is published as insufficient data rather than as a percentage.
References & data sources
- Open-Meteo ERA5 historical reanalysis — archive-api.open-meteo.com
- ESPN college football scoreboard and pickcenter endpoints
- Companion research: CFB Heat & Humidity Shock study (climate-relative, not absolute temperature) and CFB Wind & Kicking study
Model cfb-extreme-heat-v1
Frequently asked
What exactly is being tested?
A public claim that non-conference FBS games hit the Over about 59% of the time when kickoff temperature sits between 85 and 106°F, on the theory that extreme heat fatigues defenses faster than offenses. This is absolute temperature only — no team climate baselines, no travel adjustment, no acclimatisation score.
How is this different from your Heat & Humidity Shock study?
That study asks whether a visiting team is shocked relative to its own home climate. This one ignores who is playing entirely and asks only whether the thermometer reading at kickoff correlates with Over results.
Why a minimum sample threshold?
Because a hot number on a thin slice is noise, not a finding. Any temperature band with fewer than 40 graded games is labelled insufficient data and never feeds a headline — the same rule we applied when correcting our CFB wind study.
Is this a betting system?
No. It is a conditions study published exactly as the numbers came out, including when the numbers say nothing. Check the weather before you bet.