✳️The Man Trying to Improve Ultrarunning Livestreams

On the next data frontier in ultrarunning

Hey pals,

A single topic today, and tbh after the Satsify palava we all need to go outside and touch grass a little.

A lot of people asked my opinion on it and honestly, I really don’t care about it. Yes it was cringey, yes Brice and Satisfy’s attempts to quell the online fire made it worse, but let’s not make a mountain out of a molehill. If you want the story from someone who was there, read Grace’s article. If you want to know how Satisfy got here, read Seb’s piece. Now let’s move on.

Today is one of those articles that only the hardcore fans of the business side of the sport will read. I’m fully expecting one of my lowest engagement rates, but it hella interests me, so let’s do this.

Watching ultrarunning live can be both fun and frustrating. Josh has a nerdy fix for it.

Hope you all have wonderful weeks,

Matt

Josh at Borderlands Trail Running has been thinking about what’s missing from ultra running coverage and recently decided to build something. LoC (Live on Course) is a live race intelligence system that, in theory (it’s not live yet) uses human observers positioned throughout a course to capture the moments that cameras and GPS trackers tend to miss, such as comments from athletes or whether they look like confident or sketchy. Field reports are verified and shaped by moderators before reaching two audiences: fans following a public live wire, and commentators getting a deeper tactical feed to fuel their commentary. The idea is to make 20+-hour broadcasts with signal issues and limited footage feel alive between checkpoints, not just at them. It’s a working prototype right now, and Josh is open about the fact that he’s still working out what it becomes

Admittedly, it’s a bit full-on to look at right now…

As both a data nerd and watcher of many hours of livestream, this makes sense to me. Data created by field observers is part of all live televised sports. Entire companies are built around their ability to have humans report and codify whenever something happens in a game to generate the datasets that underpin the information architecture of modern sports. Today more data is being captured through optical tracking in-camera, automatically calculating data as it happens, but people simply watching a game and counting when a pass happens, for instance, is still fundamental to the business.

Josh’s idea here has a genuine use case for ultrarunning, so I thought I’d ask him some questions to get the background.

What’s the core problem you’re solving here - a livestream problem, an audience problem, or a data problem, because they each create different solutions?

I’d frame it primarily as a race visibility problem, which then becomes an audience retention problem.

In our episode Cocodona Has a Fan Problem, one of the big points I made was that the issue with longform ultra coverage often isn’t top-of-funnel interest. People enter the coverage constantly. The problem is exits.

Every 20 seconds a viewer is subconsciously asking if this is worth their time right now?

A lot of ultra coverage struggles because huge portions of the race become invisible between aid stations and drones. Cameras only catch fragments. Commentary is often bridging gaps with incomplete information.

LoC is an attempt to restore visibility to the middle of the race through distributed human observation.

If the race becomes more understandable moment-to-moment, viewers stay longer because there’s more narrative tension, more emotional continuity, and more meaningful insight.

So yes, it touches livestreaming, audience retention, commentary quality, production value, and operational awareness, but I don’t really think of it as a “data business” problem.

The goal isn’t more data. It’s better race understanding.

A huge guiding principle for us while building this has been: miss the coverage, miss the race.

Not because the coverage becomes louder, but because the insight becomes deeper.

A huge part of that is the commentator layer. The public-facing live wire is only one part of the system. The commentator feed is intentionally deeper and more operational. It surfaces developing tactical reads, emotional shifts, race dynamics, and emerging storylines that help commentary become more insightful and more connected to what’s actually happening out on course.

Is there an ambition to make this into the data business of trail running?

Not really, at least not how I think that phrase is normally used.

I’m less interested in building a giant statistics layer and more interested in building a live race intelligence layer that creates a fan experience my kids would actually join me in consuming because it’s more legible and emotionally understandable.

I want more visibility, more tactical understanding, more emotion, more storytelling from moments we’ve never really had access to before.

All of that makes the coverage harder to leave.

A huge amount of ultra racing currently disappears in real time. That’s the thing I’m interested in solving.

You mention you built this with AI. UTMB and Aravaipa already have armies of volunteers on course with phones. What stops them from building a WhatsApp channel with a good moderator and doing 70% of this for free?

As a fan, I wish they would build something that did that.

The human network is already there. Volunteers, crews, aid stations, camera ops, spectators, people scattered all over the course observing meaningful things.

What doesn’t really exist yet is infrastructure that structures those observations into something operationally useful.

A WhatsApp thread can move messages around, but it doesn’t really create moderation systems, confidence structures, role-specific routing, public vs commentator intelligence layers, structured field reporting, signal prioritization, or operational UX built specifically for live ultra coverage.

AI is mostly helping us accelerate moderation, structure, routing, and synthesis. The value still fundamentally starts with humans on course paying attention.

In some ways LoC is less about inventing a new human network and more about making the existing one legible.

Still very early obviously, but that’s the direction.

Even if you’re delivering qualitative signals, they will need some form of categorisation and hierarchy to make the editing process simpler and standardised across field reporters, ultimately creating quantitative data in the process. Part of the moat of a sports data business is their methodology and categorisation that can be applied across multiple races, so it’s a problem you will likely come against.

The interesting thing is the project started less as “sports data” and more as frustration watching races I cared about feel too hard to follow despite massive effort from production teams and commentators. The deeper I got, the more I realized the issue is really an information architecture problem.

You’re also right that the qualitative layer eventually becomes structured data whether you intend it to or not. Part of the challenge is figuring out what deserves standardization versus what loses value once over-systematized. I suspect the moat is less raw data collection and more operational judgment around signal quality, prioritization, moderation, and race relevance.

Have you reached out to RDs, commentators and athletes for feedback? Any thoughts so far?

I’ve spoken informally with those who would have to adopt it but I chose to bring it to the market first mostly for the enjoyment of seeing what fans would think if implemented.

Do you want this to be a business? If so what do you envision the business model to be at the moment?

As for business model, still very exploratory. My instinct right now is that the primary value accrues to live coverage quality and audience retention, which likely makes races, broadcasters, and production ecosystems the eventual customer more than consumers themselves. But I’m intentionally resisting pretending the business is fully solved yet. Right now I’m more interested in proving the operational concept and seeing if the signal quality can become genuinely useful in a live environment.

---

My take on it all

Josh is being modest about what he’s built, and slightly evasive about the hard problems ahead. Both are understandable. LoC is a prototype built out of genuine frustration with a coverage problem, and he’s right to resist over-formalising it too early. But the sports business world has been here before, and there are lessons worth surfacing.

When Josh says his goal is “better race understanding, not more data,” he’s drawing a distinction that sounds meaningful but will eventually collapse. Every sports data business started with a qualitative ambition. Opta, one of the original sports data companies built in the 90s, didn’t set out to build a statistics empire, they set out to help football clubs understand matches better and help fans feel closer to the game. The categorisation, the taxonomy, the structured methodology came later, out of necessity, because qualitative signal is only as useful as your ability to route, compare, and act on it consistently. Josh acknowledges this himself when he talks about information architecture, he just hasn’t fully reckoned with what that means operationally.

The WhatsApp question is the right one to pressure-test. His answer is correct, a channel moves messages, a system creates structure. Top marks. But it slightly sidesteps the harder point. UTMB and Aravaipa don’t just have people aimlessly throwing their hand in the air whenever a call out for help is made. They have relationships with those volunteers, credentialing authority over who gets on course, and every incentive to keep data value inside their own ecosystem once they realise it exists. The sports data industry’s darkest recent trend has been leagues taking data rights in-house. The NBA, the Premier League, and others have progressively squeezed third-party data partners by asserting ownership over what happens on their property. Ultra running isn’t there yet, but if LoC proves the concept, the risk is that it proves it for the race organisations rather than to them.

That all being said, Josh is genuinely onto something. ultra running is underserved in a way that most sports stopped being twenty years ago. The broadcast infrastructure is still being invented. The race organisations are run by people who care about the sport more than the data rights, at least for now. And the fan base has a demonstrated appetite for depth that most sports would envy. Think about it, these are people who will watch a GPS dot move across a map for twelve hours if the narrative is compelling enough.

What Josh has stumbled onto, almost despite himself, is that the qualitative layer of sport is the last frontier of ultra running intelligence that hasn’t been seriously captured. Every other layer has been quantified, automated, and licensed through ITRA and UTMB. The felt experience of a race in real time, routed to the people who need it, remains genuinely unclaimed territory.