How do warm and cool races compare?
Conditions are part of the race. Compare cooler and warmer editions while keeping each race day equally weighted.
What the data shows
There are 4 temperature bands with enough edition coverage for this comparison.
Below 10°C: -0.4% average of edition medians (55 editions); 10–14.9°C: 0.5% average of edition medians (74 editions); 15–19.9°C: 2.3% average of edition medians (16 editions); 20°C or warmer: 9.5% average of edition medians (4 editions). Values compare finishes with earlier recorded bests, not the visitor’s chosen target. Every edition has equal weight.
All courses · all ages · all recorded categories · all earlier-time bands. Each group shows its own sample size.
Explore the comparison
Performance relative to earlier best, by start-hour temperature
Finish-time change (%) · below zero is faster
At least three editions and 100 finishes per band. This does not isolate a temperature penalty.
View exact values and sample sizes
| Group | Performance relative to earlier best, by start-hour temperature | Observations |
|---|---|---|
| Below 10°C | -0.4% | 233,008 |
| 10–14.9°C | 0.5% | 265,822 |
| 15–19.9°C | 2.3% | 24,632 |
| 20°C or warmer | 9.5% | 754 |
How to read this
Each bar compares results with earlier recorded bests. Positive means slower. Each race edition contributes equally, regardless of field size.
Keep in mind
Start-hour temperature is an estimate. Wind, humidity, sunshine, fitness and course differences are not isolated.
How we calculated this
Use the supplied modeled start-hour temperature and runners with earlier benchmarks. Compute edition-level median finish-time change, requiring 20 finishes per edition/group. Average edition medians equally within temperature bands, requiring three editions and 100 finishes. Weather exposure is a start-hour proxy, not personal exposure; humidity, sun, wind and route changes are not isolated.
Runner comparison groups contain at least 100 eligible observations. Broader comparisons are labeled; a finish can belong to a runner with several races. Elevation describes the supplied course profile.
We check the raw timings and apply reviewed source-quality rules before forming these groups. Editions with known invalid split grids, incomplete ingestion, unresolved source checks or a selected field are excluded. Missing age or recorded gender alone does not remove an otherwise eligible finish from the overall comparison.
Input data: Marathon results, weather and course data. Read the full methods in source units or browse the marathons, years and data fields.