UTMB training: the race is won and lost on the descents

UTMB is a loop of the Mont-Blanc massif from Chamonix, through Italy and Switzerland and back: about 174 km with roughly 9,700 m of ascent. The number people quote is the climbing. The number that ends their race is the descending, which is exactly as much. Ten major climbs, ten long descents, most of it above the treeline, across two nights.

Climbs cost fitness, descents cost muscle

Going up is a cardiovascular and strength problem, and it's the one most people train, because it's obvious and it hurts in a familiar way. Going down is a tissue problem. On every descending step your quads work as brakes, producing force while lengthening. That eccentric load is what tears muscle, and the bill arrives hours later. A runner can feel strong at halfway and still come apart later, not because the engine failed but because the legs stopped absorbing landings.

Courmayeur to Champex, where the field thins

LiveTrail, the race's timing system, publishes where every runner who abandons was last recorded. From 2023 to 2026, between 50% and 57% of each year's abandons were recorded between Courmayeur, around km 83, and Champex-Lac, around km 130. Courmayeur, Arnouvaz and La Fouly were among the four biggest stopping points every year. For runners near the cut-offs, that stretch starts on Saturday afternoon and runs into the second night. From Arnouvaz the course climbs about 760 m to the Grand Col Ferret, then drops into Switzerland past La Fouly. One runner's recap says that after the col, "during the next 20k’s things started to fall apart for me". The timing shows where people stop. It doesn't say why, and a stop at Courmayeur can be a decision made on the climb before.

Building legs that still descend at hour 30

Descent tolerance has one reliable friend: the repeated-bout effect. One hard downhill session protects you from the next; the soreness after the second drops sharply. So the descents get trained early and gradually, learned fresh before they're practised tired, and backed by heavy strength work in the gym, which builds the same tissue the drops attack. Later in a build, descents go on tired legs on purpose, because that's the race. Chasing a single monster downhill three weeks out, on legs that can't yet take it, buys an injury instead of an adaptation. The full progression is in the descent training field note.

Climbing, poles and the long grind

Most of UTMB's uphill is power-hiked, not run, so strong efficient hiking is a trained skill, not a fallback. Poles are allowed and worth it if you'll actually practise with them: they take a measurable share of load off the legs on the climbs and steady you on technical ground, but only if the technique is grooved and your upper back can hold them up at hour 30. The engine underneath all of it is still ordinary capped easy aerobic volume, built long before the mountain-specific block. Vertical and technical terrain are the specificity; the base is the base.

Fuelling and the two nights

You'll be moving for the best part of two days and through two nights, and appetite fades exactly when the climbing demands most. That makes fuelling a rehearsed skill: a target of grams per hour you can absorb on steep ground, tested on long mountain days, with food you'll still tolerate cold and tired in the dark. Add the temperature swing, valleys near 30°C, cols below freezing, and pacing that respects both, and the race stops being a fitness test and becomes a management test.

Where durability shows up

UTMB is the clearest case of the thing Durability is built around: your aerobic system can be intact while your legs stop delivering. On a long mountain run the signal I read is downhill pace late in the run against the same descents fresh, how much speed disappears after a big climb, and whether foot control and braking survive once the legs are tired. Be clear on what the watch does here: grade-adjusted pace and heart rate on steep technical ground are approximate, so I read the durability numbers on the runnable, comparable sections and treat the technical descents as observed evidence, logged, not measured to two decimal places. The data points me at the weak link. The decision to load descents, add gym work or hold volume is mine.

Thirty weeks for a late-August start

I build UTMB over 30 weeks from late January: capped easy volume and two heavy strength sessions a week, kept 48 hours clear of the key runs, then the climbing, then the Alps.

• Downhill repeats from week one: 3 × 5 minutes at 5 to 10% gradient, at least every two weeks. Then longer, steeper, and later in long runs.
• The Col Ferret day: a long run built so the biggest climb and the longest descent come after six hours on your feet.
• Mountain back-to-backs every two to three weeks in the alpine block: about five hours on Saturday, three to four on Sunday, with poles.
• Pole power-hiking intervals on the steepest hill you can reach.
• Long runs that finish after dark, with grams per hour logged on the climbs.

The biggest back-to-back lands about five weeks out. If you live somewhere flat, treadmills and stairs carry the climbing, but the descents have to be found on real ground. Then a three-week taper, because mountain volume takes more out of the legs than it looks on paper.

What I watch

• Downhill pace late in long runs, against the same descents run fresh.
• How much speed the legs lose after a sustained climb.
• Whether braking control and foot placement hold once tired, or go ragged.
• Climbing efficiency with and without poles, and whether the upper body lasts.
• Fuelling on steep ground: grams per hour logged, gut score out of ten.
• Recovery after the biggest hill weekends, because that's where masters legs get caught.

Facts checked September 2026 against the race's 2026 information: about 174 km and 9,700 m of ascent, roughly a 46-hour limit, drop bags at Courmayeur and Champex-Lac, no pacers, entry via UTMB Index plus Running Stones into a lottery. The 2025 and 2026 courses ran slightly short after a technical col was bypassed; 2027 dates were not yet published. Where runners stop: my count, October 2026, from LiveTrail's abandon statistics for the UTMB (utmb.livetrail.net for 2026, livetrail.net/histo for 2023 to 2025). Abandons recorded from Courmayeur to Champex-Lac: 533 of 930 (2023), 554 of 1,001 (2024), 412 of 827 (2025) and 410 of 725 (2026). Kilometre points, the 760 m climb and the back-of-field timings are from the course data on the organiser's race page, which may not match the 2027 course exactly. The quote is from a runner's published race recap. The build numbers are my coaching method: downhill repeats of 3 × 5 minutes at 5 to 10% at least every two weeks, back-to-backs of about five hours plus three to four, and a three-week taper for a 100-miler. Evidence level: eccentric damage, the repeated-bout effect and pole load reduction are well supported; the training progression is coaching practice built on that evidence.