Across 1.79 million marathon performances the air temperature for the fastest running was 3.8–9.9 °C depending on level; performance fell on both sides, and withdrawals rose with heat
Main caveat: Race-level temperature against the speed distribution, with modest fits (r² 0.24–0.27 in the examples shown); wind was not analysed.
Protocol
Design: observational analysis of race results against weather and air quality.
Data: 1,791,972 performances (every finisher per race and year) from six marathons — Paris, London, Berlin, Boston, Chicago and New York — from 2001 to 2010, 60 races in all.
Exposures: air temperature, humidity, dew point and sea-level atmospheric pressure for each race, and the concentrations of NO2, SO2, O3 and PM10.
Outcomes: the distribution of performances in each race (mean and SD) and its relation to each environmental factor (a quadratic model for temperature); withdrawal rates.
Key Findings
The optimal air temperature for maximal running speed varied from 3.8 °C to 9.9 °C according to performance level.
Six marathons (Paris, London, Berlin, Boston, Chicago, New York) from 2001 to 2010.
Large decreases in temperature below the optimum also reduced performance.
Withdrawals rose with temperature; Chicago 2007 had 30.74% withdrawals at a race temperature of 25 °C.
Practical Application
A race day of 4–10 °C is close to ideal for a marathon, not "cold". At 12 °C a hard effort is already on the warm side of optimal.
So for racing in the cool, the mistake to avoid is overdressing, not underdressing — see SYNTH-79.
Limitations
Race-level temperature against the speed distribution, with modest fits (r² 0.24–0.27 in the examples shown); wind was not analysed. Full text was read.
How this entry was checked and changed
Content correction 2026-10-27
Tier line rewritten from the entry's own participant description (McKay 2022 bands; a band marked "inferred" was not stated by the authors); checked by an independent second reader.