Temperature, Wind, and Domes: A Statistical Breakdown of Weather and College Football Scoring

Analysis · by Rick's Picks Analytics

Building on our six-hypothesis weather study, this analysis takes a more granular approach -- segmenting games into precise temperature bins and using ANOVA to detect scoring differences across conditions.

Temperature ANOVA: Five Climate Zones

Instead of simple hot-vs-cold comparisons, we divided games into five temperature bins: Freezing (below 32F), Cold (32-50F), Cool (50-70F), Warm (70-85F), and Hot (above 85F). We then ran a one-way ANOVA to test whether mean total scoring differs significantly across these five groups.

Why ANOVA? A t-test only compares two groups. ANOVA tests whether ANY of the five group means differ from each other, protecting against the multiple-comparison problem of running 10 separate pairwise t-tests.

What the data shows: The ANOVA reveals whether temperature, treated as a categorical variable across five bins, has a statistically significant association with total scoring. Post-hoc pairwise comparisons (with Bonferroni correction) then identify which specific pairs of temperature bins differ.

The key finding: The biggest scoring gap is between Freezing (below 32F) and Warm (70-85F) games. The difference shrinks as temperature increases into Cold and Cool ranges. Hot games (above 85F) score similarly to Warm games, confirming that the weather effect on scoring is primarily a cold-weather phenomenon.

Dome vs Outdoor: Controlled Environment

We handle dome games carefully in our analysis. Rather than guessing dome temperatures, we code dome games as their own category for temperature analysis and set dome conditions to temperature=72F, wind=0 MPH when needed for weather-based models.

The t-test result: Dome games show higher mean total scoring than outdoor games. This is a consistent finding across both training (2015-2022) and test (2023-2024) periods. The Cohen's d effect size tells us whether this difference is practically meaningful or just statistically detectable due to large sample sizes.

Practical implication: The dome effect is real but modest. On its own, it might shift a projected total by 1-3 points. We incorporate it as one of several adjustments rather than treating it as a standalone betting signal.

Cold vs Warm: The Sharpest Divide

Our most targeted comparison: games below 32F versus games at or above 50F. This deliberately compares the extremes to maximize signal.

Training set (2015-2022): We compute the mean total scoring for each group, run a Welch's t-test, and measure Cohen's d.

Test set validation (2023-2024): We repeat the same analysis on held-out data. If the training-set finding replicates in the test set, we have strong evidence of a real effect rather than a statistical fluke.

What we find: The cold-weather scoring suppression is one of the most robust signals in our entire analytical framework. It replicates across time periods with consistent direction and similar magnitude.

High Wind Analysis

We compare games with winds above 20 MPH to games with winds at or below 10 MPH. This is a stricter threshold than our hypothesis study (which used 15 MPH) to focus on conditions that meaningfully impair gameplay.

The finding: Above 20 MPH, wind measurably reduces scoring. The effect is driven primarily by suppressed passing efficiency and reduced field goal accuracy. Games with moderate winds (10-20 MPH) show a weaker effect, suggesting a nonlinear relationship where wind matters most above a threshold.

How This Feeds Our Algorithm

The temperature and wind effects are combined into a single "weather adjustment" factor in our prediction system. For any upcoming game, we check the forecast and apply adjustments based on the expected conditions:

• Below 32F: Strongest downward adjustment to projected total
• 32-50F: Moderate downward adjustment
• 50-70F: No adjustment (baseline conditions)
• 70-85F: No adjustment
• Above 85F: No adjustment
• Dome: Slight upward adjustment
• Wind above 20 MPH: Additional downward adjustment, stacks with temperature

These adjustments are calibrated from the training data magnitudes and validated for stability on the test set.

Analysis uses one-way ANOVA with Bonferroni-corrected post-hoc pairwise comparisons. Temperature and wind effects validated on 2023-2024 holdout data.

Rick's Picks publishes statistical analysis of college football for informational and entertainment purposes. Nothing here is betting advice. 21+. If gambling is affecting you or someone you know, call 1-800-GAMBLER.