Eleven Editions of the Alpe d’Huez Triathlon: 31,140 Finishes, Three Races, One Mountain

I spent a weekend importing every result the Triathlon de l’Alpe d’Huez has published since 2015 — 11 editions, 33 races, 31,140 classified finishes out of 35,579 entries. 2020 is missing because 2020 did not happen.

Three races run on that weekend — the Triathlon L, the Triathlon M and a Duathlon — and all three are in here. What makes it worth an article rather than a database is that the long course carries the same five legs in all eleven editions: swim, T1, bike, T2, run. Eleven years of the same lake, the same valley and the same 21 hairpins, measured the same way. That is a rare thing in amateur sport, where results usually evaporate into a PDF within about eighteen months.

Everything below comes from the files of ChronoRace, the event’s timekeeper. All of it is browsable, edition by edition, in the race explorers — including your own race, if you were on that mountain. I was not: I have ridden up the Alpe, but never after swimming across a lake first, which on this evidence was the correct life choice.

What is actually in the archive

Three races run on the same weekend: the Triathlon L (the long one, and the one people mean when they say they did the Alpe), the Triathlon M, and a Duathlon. All three have grown, but not equally — the Duathlon has more than doubled while the M has barely moved since 2024.

Worth pausing on which is the big one, because it is not the one people talk about: across eleven editions the Triathlon M has 13,986 finishers to the L’s 12,078. The Duathlon adds 5,076. The L gets the attention and the M gets the entries.

The next three sections take the L first — it has the cleanest data by some distance — then the M, then the Duathlon. After that the mountain lets me compare all three directly.

The Triathlon L: the front got faster, the middle did not

The winning time has come down by more than twenty minutes: 5:55:14 in 2015, and 5:30:05 in 2025 — the fastest edition in the archive. The median finisher has done nothing of the sort. The middle of this race is not getting quicker; some of the slowest median years are recent ones.

One honest caveat before anyone screenshots that chart: the organiser did not publish course distances before 2026, so I cannot prove the bike leg was identical in every edition. Absolute finish times across years carry an unmeasurable course component. That is exactly why the two sections below exist — they lean on the parts of the day whose length is not in doubt.

So which was the best edition of the L?

Depends what you mean, but at the sharp end the answer is not close. In 2025, 21 athletes went under six hours. In 2019, nobody did. Not one.

2025 takes the fastest winner, the deepest front group and the most sub-seven-hour finishers in the archive. 2026 takes the biggest field, the highest share of women, and the fastest ascent of the mountain itself. If you want a single answer: 2025 was the fastest edition ever run here, and 2026 was the biggest.

The swim is the constant. The bike is where the day lives.

Across eleven editions the median swim ranges from 42:16 (2017) to 46:04 (2015) — a spread of about 4.3% either side of the middle. A course that moved would not produce a band that tight. The swim is my yardstick, and it is the reason I trust the next observation: while the swim held still, the median bike leg wandered by nearly three quarters of an hour, from 5h27 in 2015 to 6h12 in 2019.

The transitions deserve a footnote of shame. The median T2 has gone from 155 seconds in 2015 to 224 seconds in 2026. We are, collectively, getting slower at putting on trainers, and I refuse to believe this is a course-design issue.

The Triathlon M: the biggest race, and the one that changed most

The M is the short course and the crowded one — 13,986 finishers across eleven editions. Two things separate it from the L, and both are large.

Its winning time has not improved. The fastest M in the archive is 1:47:13 in 2022; the 2026 winner needed 1:55:18, which is slower than 2015. Where the L’s front end sharpened over a decade, the M’s did not move at all — and its median went the other way, from 2:47:10 to 3:26:27.

And it is where the women are. Women were 16% of M finishers in 2015 and 32.5% in 2026 — nearly a third of the field, against 14.3% in the L. If you want to know where this sport is actually growing, it is not the hero distance.

There is a catch, and it is the reason the M gets no cross-year time claims from me. Run the same yardstick I used on the L — the swim — and it fails badly: the M’s median swim goes from 23:56 in 2015 to 36:16 in 2026, a drift of 52%. The L’s stayed inside 4.3% either side of its middle over the same decade.

The Duathlon: run, bike, run — and one man

The smallest race and the fastest-growing: 317 finishers in 2015, 769 in 2026, a factor of 2.4. No swim — you run, you ride up the Alpe, you run again.

And then there is Émile Blondel-Hermant, who has won it 6 times, including the last 5 in a row. He also turns up in the repeat-finisher list with nine editions, and he was second in the 2026 Triathlon L. The Duathlon is, on this evidence, his.

Its splits are the thinnest in the archive, so I will not push them far: 2021 publishes no segment results at all, 2018’s sheet merges legs together, and the closing run roughly doubled between 2019 and 2022 — 14:11 to 27:53. That is a course change, so nothing here compares a Duathlon run across that line.

The 21 hairpins: the one segment nobody can argue about

Four editions publish the organiser’s own split for the montée de l’Alpe — the final climb of the bike leg, 13.4 km of the 121.8 km course in 2026. I checked it really is the climb and not something else: in all four editions it accounts for 25–29% of each athlete’s own bike time, which is what a famous hill at the end of a long flat valley does to a split.

The fastest ascent improved from 51:15 in 2016 to 48:11 in 2026 — three minutes off the climb, after a hundred kilometres of racing. The median ascent, over the same decade, went precisely nowhere: about an hour and a half, then and now. It is the whole thesis of this article in one segment.

The climb is also the only place where the three races can be measured against each other, because it is the same 13.4 km whichever event you entered. In 2016, 2018, 2026 all three publish it.

In 2016 and 2018 the M rode the climb the way you would expect — several minutes quicker than the L, because its riders arrived with far less race in their legs. By 2026 that advantage is gone: the two medians land within three seconds of each other, 1h31 against 1h31. The climb did not change. The M’s field did — it is now a much broader race than it was, and the mountain reports that faithfully.

The fastest first name in France, and why that is a lie

31,140 finishes carry 4,037 distinct first names. The most common one, by a distance, is Nicolas — 539 finishes. I want to be clear that I did not rig this. I have never entered this race, and neither has any other Nicolas Sitter: the archive contains 0 of us.

Martin tops the surnames, which is the least surprising fact in this article. Further down sits Chevalier (41), a name that also won the thing in 2022.

Now the fun one. Take every first name with at least 25 finishes on the long course and rank them by median time, and you get a leaderboard of fast names and slow names. Victor is the fastest name in the archive (7:57:29 median); Alain is the slowest (9:58:55). That is a gap of more than two hours, purely on the basis of what your parents chose.

It is, of course, entirely an age chart. Plot each name’s median finish against the mean age band of everyone carrying it, and the joke collapses into a straight line:

Which means the name data is really a demographic instrument. If you count two fixed baskets of names — one that peaked in France around the 1950s and 60s, one that peaked in the 1990s — you can watch the sport hand itself over to a different generation, and pin the handover to a single edition.

Two more, for no reason at all: 1,694 finishers share an initial between their first and last name (5.4% — about what you would expect by chance), and 525 carry a hyphenated first name, the Jean-Pierre tax.

Who shows up

The long course has grown from 883 finishers to 1,559, and its women’s count from 77 to 223 — nearly tripling while the field itself did not quite double. As a share that is 8.8% to 14.3%, a gain of 5.5 points, and still less than half what the M manages.

The field is not ageing, which surprised me. Across the six editions whose category codes carry an explicit age band, the mean sits between 37.4 and 39 and goes nowhere. What has happened is that both ends grew: under-35s from 28.1% to 38.2%, and the over-50s from 19.2% to 24.3%. The middle thinned.

It is also an international race that stays stubbornly French: between 56.7% and 66.1% of finishers are French in every edition I can measure, while the number of countries represented has climbed from 29 to 46. Belgium has quietly become the biggest visiting nation, overtaking Great Britain around 2024.

Every winner of the L and the M, 2015–2026

The women’s race has been the more interesting one to follow. Alanis Siffert has won the last 3, and Daniela Ryf in 2019 finished 6th overall — the highest any woman has placed in the mixed field in this archive.

The M has been harder to repeat: only Mikel Arrassate (2 wins) and, in the women’s race, Justine Guérard and Audrey Merle (two each) have won it more than once. Its highest female finish overall is 11th, by Andrea Hewitt in 2016.

The Duathlon’s winners are in its own section above. Nationalities are as the timekeeper recorded them, and 2021 and 2023 published none at all, hence the blanks.

Two people have finished all eleven

There is no athlete id in these files, so “the same person” means the same accent-folded name. That is a real assumption and I have measured what it costs — see the limits section. But at the top of the distribution the assumption is doing very little work, because a name would have to be shared by an implausible number of humans to fake eleven appearances.

Franck Dumenil and Hugues Peyret have finished every edition since 2015. Dumenil did it the hard way — the long course, all eleven times. Each has an independent corroboration built into the data: their age category marches upward exactly as one human ageing eleven years should, from Veteran 2 to M60–64 in Dumenil’s case, across five different category vocabularies that the timekeeper changed under them.

He was 553th of 883 in 2015, and 1,383th of 1,559 in 2026 — from the middle of the field to the back third, while the field itself nearly doubled and he went from his late forties to his sixties. Two hours slower, eleven for eleven. I find that a much better result than the winning time.

Beneath them, 4,479 people have finished more than one edition, and one woman — Vanessa Vidal — has nine.

Where the long course breaks people

Non-finishers only leave a trace if they crossed at least one mat, and in the older editions most of them did not: in 2015, 94 of 233 did. So this is restricted to the four editions where at least 85% of non-finishers left splits behind.

There is no single answer, which is itself the finding: 2024 and 2025 lost most of their DNFs on the run, 2026 lost more on the bike. This is a race with two entirely different ways to end your day, and which one gets you appears to depend on the weather more than on the course.

What this data cannot tell you

Eleven editions come from twelve different spreadsheet layouts and two timing providers. Here is everything I know to be soft, so you can discount the right claims rather than all of them:

• No course distances before 2026. Only the 2026 edition publishes leg lengths. Nothing here converts an old time into a speed, and cross-year time comparisons carry a course component I cannot size.
• 2017 and 2022 publish finishers only. Their exports contain no DNF rows at all, so their finish rate is 100% by construction. I have kept them out of every attrition number.
• I do not publish exact race dates for the spreadsheet editions. The dates in my database for 2015–2024 were derived from the year as a placeholder. The timekeeper never told me those dates, so quoting them would be inventing a fact.
• Nationality is incomplete. 2023 publishes none; 2021 publishes it for 446 of its 2,731 finishers. Both are excluded from the country series rather than drawn as a collapse. Codes also changed standard mid-archive (NED became NLD, GER became DEU), and I normalised them.
• Repeat athletes are name matches, and I have priced the error. Two people in the same race cannot be the same human, so collisions within a single race measure how often two athletes share a folded name: 15 times in 31,120 rows. Applied to every cross-race pair, that predicts roughly 401 false merges across the whole archive. So treat the “2+ editions” buckets as an upper bound: that budget is about 13% of the two-edition group, so perhaps one in eight of those pairs is really two different people with the same name. The eleven- and ten-edition names are a different matter: each is corroborated individually by its age-category progression.
• The run has not been quite the same course throughout. From the 16th edition in 2022, the run leaves the Col de Sarenne road and descends the airfield’s takeoff runway before rejoining the old route, instead of dropping to the football stadium. The organiser publishes the lap as unchanged — 6.5 km and 110 m of climbing — and the split data shows no step across 2021→2022, so nothing here is adjusted for it. But a 2015 run split and a 2026 one were not run on quite the same ground.
• The M’s swim moved and the Duathlon’s closing run doubled. Both show up as step changes in the leg medians, and neither is explained by anything the organiser published. The consequence is that the M and the Duathlon get no cross-year time claims in this article, only within-edition ones and the climb.
• The Duathlon has the thinnest splits. 2021 publishes no segment results at all, and 2018’s sheet merges legs together (run-1 with T1, bike with T2), so those two editions are absent from its leg table.
• One whole column was not what it claimed. The Triathlon M’s 2015 “montée” has a median of 8:48 and all 1,052 of its values fall under a 40-minute floor. It is excluded rather than plotted. That is a plausibility floor earning its keep.
• Age bands exist for six editions. The 2015–2018 and 2021 category codes (MS1…MV4, S3H, V4H) encode age indirectly. I did not convert them rather than guess, which is why the age section is shorter than the others.
• One corrupt split survives upstream. A 601-second bike leg in 2018 is obviously not a bike leg; it is excluded from best-time figures. Across 31,140 finishes that is the only leg-level corruption I found after the plausibility checks — which is a good advertisement for the timekeeper.

If you want the gory version, the import notes for the older editions live in the repo alongside the scripts, including the edition that stored every time as an Excel day fraction and briefly convinced my validator that 2,836 people had finished a triathlon in no time at all.

Browse any edition

Every edition has its own explorer: leaderboards, leg splits, categories, clubs, first names, and a page for every athlete.

If you raced any of these — L, M or Duathlon — your own eleven-year arc is in there. Franck Dumenil’s just happens to be the longest, and Émile Blondel-Hermant’s the most decorated.