College Football Over/Under Totals

Conditions-adjusted total

Pick the conditions and the tool reports where totals have historically closed in them. The controls are bands rather than free numbers because the model is measured in bands - there is no coefficient that resolves 17.3 mph differently from 16.8 mph, and a box that accepted one would imply there was.

What actually moves a college football total

Weather acts on the total far more than on the spread, and the reason is structural: wind degrades both offences at once. The passing game and the kicking game suffer first, which lowers the points expected in the game without changing which team is better. A forecast can therefore be dramatic and leave the spread untouched.

Everything below was measured across 3,138 non-neutral FBS-v-FBS games from 2021-2025, against an average closing total of 53.6 points. Three columns are reported for every band, and they answer different questions:

• Closing total vs average - where the total actually landed in these conditions. This is what the tool above applies.
• Open-to-close move - how far the market itself moved the number before kickoff. This is the market's own behaviour, not an opportunity.
• Left after close - the over/under result relative to the closing number. This is the only column that could indicate an edge, and it is small, noisy, and on the windiest band it points the other way.

Sustained wind

Gusts

Temperature

What this model will not tell you

Stating the limits is part of the product, not a disclaimer bolted to it.

Rain is not modelled, and cannot be

The precipitation field in the underlying data is a single hourly reading. Across the sample, 88% of games recorded exactly zero, with a mean of 0.08 mm and a maximum of 8.7 mm. There is not enough variation there to detect a rain effect of any size, in either direction. A "rain does not move the number" result would be a statement about the measurement, not about football, so no rain coefficient is published. Recording game-total precipitation, or the archived forecast the market actually priced, would be needed to answer the question properly.

Bands with too little evidence are shown but not applied

A band measured on fewer than 30 games is reported for completeness and excluded from the calculation: 20+ mph (3 games); below 32°f (freezing) (28 games). The 20+ mph wind figure is the clearest case - it reads -7.23 points, which would be by far the largest effect on the page, on the evidence of 3 games. The freezing band sits just under the threshold at 28 games; the cut is a round number chosen in advance rather than one tuned until the answer looked right.

Factors with no measured totals effect

Blue Chip Analytics measures precipitation, humidity, pace, offensive efficiency, defensive efficiency, rest, travel, elevation against the spread, and several of them carry real charges there - see the travel and line value pages. None of them has a measured effect on the total, so none is applied here. They are named rather than quietly omitted, so that their absence reads as a known gap rather than an oversight.

Two conditions are added as though independent

Wind and temperature are measured separately, each against the same overall average. Adding them assumes they are independent, and they are not perfectly so - cold games tend to be windier. Sustained wind and gusts are not added to each other for the same reason in a stronger form: they describe one physical condition, so the tool applies whichever is larger and reports the other for context.

The weather is what happened, not what was forecast

The measurement uses realised conditions, while the market priced a forecast. These are therefore plausibility bands rather than a reconstruction of what the bookmaker knew.