What heat does to an Ironman marathon

The question

Writing up IRONMAN Florida 2026 we noticed that the same course, on a day about 9 °C warmer in 2024, had a median age-group marathon half an hour slower. One course, three years, is an anecdote. We hold the official results of every full-distance IRONMAN since 2011 that we have imported — 203,221 results from 110 race days — and a race-day weather record for 97 of them, from 13.3 °C (Wales 2025) to 34.9 °C (Vitoria-Gasteiz 2026). So: what does heat actually do to an Ironman, and to which part of it?

Two ways to look at it. Across race days, we compare each field’s median marathon as a share of its median bike split against the day’s high: the bike is the control, so a hilly or slow course doesn’t masquerade as a hot one. Then the stronger test: the same athletes on different days. 24,889 age-groupers in our results have finished two or more full-distance races; for each we take their coolest and their hottest race and compare the splits. No names appear anywhere in what follows; everything is an aggregate.

The same athletes, a hotter day

Take every age-grouper with at least two full-distance finishes in our results, line up their coolest race against their hottest, and the pattern is unambiguous. For athletes whose two days were within 3 °C of each other the marathon moved by +0.4% — noise. At 6–10 °C warmer it was +3.2%; at 10 °C or more warmer, +5.8%, with 71.1% of athletes slower. The bike barely registered: +1.0% for the same athletes on the same days. The swim moved +2.0%, but water temperature, wetsuit rulings and currents vary far more between venues than air does, so read nothing into it. Fitted across all 17,763 pairs, a degree of race-day high is worth 0.53% of marathon time and 0.05% of bike time.

Why the bike escapes: airflow. At 35–40 km/h the body sheds heat it cannot shed at 10 km/h, and the bike comes earlier in the day, before the afternoon high. The marathon is run in the hottest hours, at the speed where convective cooling is weakest, with the day’s dehydration and glycogen debt already banked. The split that already costs age-groupers the most time is the one heat attacks.

82 race days

The same thing shows at the level of whole fields. Plot each race day’s high against the field’s median run-to-bike ratio and the line slopes up: 0.0047 per °C, or 7.4% more marathon relative to the bike for every 10 °C (r = 0.38 — courses, fields and wind account for the rest). The lowest ratios belong to Lanzarote 2025 (0.607, 22 °C) and Lanzarote 2026 (0.637, 24.4 °C), where a mountainous bike stretches the denominator; the highest to Maryland 2026 (0.924, 26.5 °C).

The band table has a wrinkle worth reading honestly: the 30 °C-and-over days (10 of them) show a quicker median marathon (4:39) than the 26–30 °C days (5:03). Those are mostly European heat-wave editions — Frankfurt, Vitoria-Gasteiz, Klagenfurt — with fast, deep fields on fast courses; the 26–30 °C band is heavy with humid North American and Asian races. That is exactly why the same-athlete comparison matters: it removes the course and the field, and it gives the same answer every time.

Same course, different years

The nearest thing to a natural experiment is a course we hold in several years with different weather. Vitoria-Gasteiz ranged 12.8 °C across its three editions; Switzerland Thun ranged 11.6 °C across its three editions; Kalmar-Sweden ranged 7.6 °C across its three editions. Read honestly, the course view is noisier than the athlete view: in five of these six courses the hottest edition also had the longest median marathon (Vitoria-Gasteiz, Kalmar-Sweden, Gurye Korea, Wales, Wisconsin), but Switzerland Thun’s hottest day (31.5 °C) was its quickest. Different years bring different fields and different winds; a hot forecast also keeps some of the less-prepared at home. That is why the same-athlete comparison, which carries the field with it, is the number to trust.

Who pays the heat tax

• Everyone, in proportion. Men’s marathons lengthen 5.3% per 10 °C in the paired comparison, women’s 5.0%. Across race days, the fastest fifth of each field and the slowest fifth slow by almost the same share of their bike split (6.0 vs 5.6 per mille per °C).
• Drop-outs climb, modestly. Across race days the drop-out rate rises about 2.5 points per 10 °C, and the share of finishers running over five hours about 9.6 points. Interestingly, the morning temperature predicts drop-outs better than the afternoon high (r = 0.45 vs 0.32): a warm start means a warm swim exit and a bike with no cool hours in it.
• Humidity didn’t show in this data. Afternoon relative humidity has no correlation with the run-to-bike ratio across race days (r = -0.04); apparent temperature does marginally better than the plain high (r = 0.41 vs 0.38). The reanalysis is a grid-cell average, not the course, so treat that as “not detectable here”, not “doesn’t matter”.

The thing that predicts a DNF better than the weather

While we had every race day open, one more thing. Rank each field by swim time and split it into quarters. Across all 97 race days, 14.3% of the slowest swim quartile did not finish, against 4.6% of the middle half and 4.9% of the fastest quartile. The effect holds on hot days (16.1% vs 5.7%) and cool ones (17.1% vs 4.9%), so it is not heat. It is worst in lake swims (18.7%) and least bad in sea swims (12.1%). A slow swim doesn’t break anyone by itself; it marks the athletes with the least in the bank for a ten-to-fifteen-hour day.

What to do with it

• Re-plan the marathon, not the bike. If your race-day forecast is 10 °C above what you trained in, expect about 5.3% on the run (14 minutes on a 4:30 marathon) and almost nothing on the bike. Go out at the pace the day allows; the athletes who held their bike plan and adjusted the run were the ones who matched this curve rather than falling off it.
• Fuel for the temperature, not the distance. Heat changes fluid and sodium needs before it changes carbohydrate. Every race guide on metrica carries typical race-day conditions and a forecast in race week, and the fuel planner adjusts its hourly plan to the forecast temperature.
• Treat swim fitness as insurance. The slowest swim quartile is 2.9 times as likely not to finish, in any weather. Nobody wins an Ironman in the water, but a lot of people lose one there without noticing.

Method: full-distance IRONMAN editions in our results with at least 300 age-group starters and a race-day high in the ERA5 reanalysis; shortened courses excluded (2). Race-day high is the daily maximum 2 m air temperature at the course coordinates; afternoon values are the 12:00–17:00 local mean. “The same athlete” is a name-and-country match across editions among standard age groups; the comparison takes each athlete’s coolest and hottest race — which may be years apart, so age, experience and course all move too, in both directions — and reports medians. World Championship days are plotted but excluded from the fits, and World Championship finishes are left out before pairing. Every figure is an aggregate; no athlete is named.