VMM 2026: Sa Pa, 5,300 Pairs of Legs and the Test of Numbers

Opening

3:47 a.m. on September 19, 2026. On the 161 km course of the Vietnam Mountain Marathon, the roped descent sits about four kilometres from the last aid station. The fog in Sa Pa was thick enough that a runner's headlamp could only pick out a smear of mud right in front of the shoe. Above, one runner grips the rope, pulling one step at a time. Below, the next waits his turn, hands on knees, breath steaming.

Images like this will flood social media over the next few days. People will talk about willpower. They will talk about courage. They will talk about the desire to conquer human limits. I have no argument with any of that. I only want, before we all nod along, to ask a simpler question: what actually decides the order of finish in Sa Pa this year?

Numbers whisper. Those willing to listen hear an entire race.

Context: a race that has changed rank

The Vietnam Mountain Marathon 2026 runs from September 17 to September 20, 2026, in Sa Pa, Lao Cai province. The organising committee BTC VNM 2026 reports more than 5,300 entrants from 54 countries and territories. The two headline distances are 161 km and 100 km, both starting at 8 a.m. on September 18. Champions are confirmed on September 19.

One thing needs to be stated plainly: this is no longer a purely grassroots event. VMM this year sits inside the World Trail Majors system — a group of internationally recognised mountain-trail races. A race in Vietnam's northern highlands entering that system changes how people read the results sheet. A slot in the system brings standards for the course, for safety, for timing, and for how results are published.

I once sat down to compare data from several well-known regional trail races to gauge scale. The figure of 5,300 runners places VMM among the most crowded mountain races in Southeast Asia. But scale does not equal difficulty. A crowded race can simply be one that is easy to enter, easy to afford, easy to logistically organise. Difficulty lives elsewhere — and it lives on the elevation chart.

Geography matters too. Sa Pa sits at an average elevation above 1,500 metres, with a temperate climate inside a tropical region. For international runners coming from Europe or North America, those are familiar conditions. For runners arriving from tropical countries in the region, they are an adaptation challenge. One course, two entirely different physiological equations.

The core: a sawtooth chart and an unmeasured variable

Before trusting a number, ask where it was born. The 161 km course at VMM 2026 has a sawtooth profile — up, down, up, down, repeated until the elevation chart looks like the teeth of a dull knife. For a runner, this shape is harsher than a single tall mountain, because the thigh muscles must work in two modes: pushing up on the climb, braking down on the descent. Every descent breaks the muscle in a way climbing cannot.

This year, record rain turned the entire trail system into mud. Slippery rock slopes became a lottery. Some sections were described as natural mud slides. Some required fixed ropes for runners to hold. And on some, fog cut visibility so low that a single misplaced step became genuine risk, no longer a footnote.

This is where I have to admit my own limits. In the performance models I once built for trail races, I had variables for gradient, temperature, altitude above sea level, and spacing between aid stations. I had no variable for mud. A 161 km race in dry conditions and the same course under record rain are two different problems, even though the elevation chart does not shift a single metre.

In other words: the same map, an entirely different race. That is the sentence static data never tells you, and the sentence social-media viewers routinely skip when comparing this year's records with last year's.

Let me try a rough comparison. On a sawtooth course at Sa Pa's average altitude, the finishing time of the leading group usually depends on three clusters of variables: climbing base, descending ability once the legs are fatigued, and fuelling strategy. In mud, all three shift. Climbing base loses value because every upward step slips part of the way back. Descending ability is tested twice over because friction drops. Fuelling strategy becomes a survival factor, because longer station stops mean the body cools and is harder to restart.

Notably, the organisers chose an 8 a.m. start for both distances. For the 161 km field, that means most runners will run through the night. A September night in Sa Pa is not dangerously cold, but paired with rain and mud it is a different story. Limited light makes reading the terrain harder, and every error in foot placement carries a price.

The 100 km field starts at the same hour but faces a different problem. It is shorter, so speed can be higher, and precisely for that reason pacing errors cost more. Going too fast over the first 30 km of mud is the surest way to pay for it in the last 30 km.

A note on data sourcing. What I am analysing here is compiled from the organiser's releases and published reports. When official results are released, I will cross-check again. In trail races, timing and checkpoint data tend to be updated progressively, so the early picture can differ from the final one. That is why I always record which data version I am working with.

The contrarian angle: mud is the lead character, but not the only cause

The media will fixate on the mud. That is understandable. Mud makes good images, good stories, a good sense of the harshest race in Vietnam. But correlation is not causation. The presence of mud at a race does not by itself explain who finishes ahead of whom.

I have made this mistake before. In 2026, when European football returned to empty stadiums, I ran a prediction model that priced home advantage at 0.45 goals per match. After nine rounds without crowds, the figure dropped to 0.08. I had to fix the model before fixing the article, because I had ignored a variable I assumed was obviously constant. Since then, every analysis of mine carries a small section titled assumptions that may be wrong.

In Sa Pa, the ignored variable may be something else. It may be night-running experience. It may be how to allocate water and salt when you sweat heavily while the temperature is low. It may be the ability to accept that you will move slower than planned, and not panic about it. None of that sits in any data table. It only surfaces at kilometre 120, when all the arithmetic has run dry.

One more point is worth raising: the presence of well-known figures. Actor Nhan Phuc Vinh is taking part and is regarded within artist circles as a serious trail runner. Media will enjoy that story, and it is good for the race. But a clear distinction is needed: an actor who runs well and a professional competing for placing are two different categories. Conflating them is a familiar error — assigning weight by fame rather than by performance data.

What the current data allows us to conclude

There are three things I will say, at a moderate level of confidence, based on what has been published so far.

First, the scale of 5,300 runners from 54 countries speaks to the race's appeal, not to the course's difficulty. Do not confuse the two.

Second, record rain turns VMM 2026 into a different version of itself. Any comparison with previous years' records needs a caveat about race conditions. A season missing detail is like a match missing stoppage time: you know the result, but you do not know how it was produced.

Third, membership of the World Trail Majors system raises the bar for organisation. In bad weather, the safety layer — ropes, aid stations, visibility limits — becomes the most scrutinised element. That deserves recognition, even when it produces no beautiful photograph.

What to track next

The VMM 2026 champions are confirmed on September 19. When the results land, I will do one simple thing: plot finishing times onto the same chart as the elevation profile, and look at how pacing was distributed. Which sections the winner ran slower than rivals, and where they made it back. Mud does not lie, but mud does not tell the whole story either.

Analysing the wrong variable is like losing your bearings for a whole year. I would rather wait a few more days of data to speak accurately than speak loudly today.