The Slush Puffer, and Other Stories from the Forest

A note from Angus:

Hello everyone! So, as part of my nerding into the data behind the Puffer, I've been analysing historic results over the last 16 years. Looking for interesting patterns.

And here’s me thinking I was alone in this strange obsession...

With an amazing (and honestly quite surprising) amount of people getting onboard with the Pufferized app over the last couple of months, I've had some brilliantly deep chats with fellow nerds (in the most positive sense of the word ‘nerd’).

It seems I am not alone in being fascinated by this event, and it having taken a very special place on my calendar.

Technical preparation aside, it’s amazing how what could appear to be such a simple (if objectively absurd) interest - in cycling in loops around a forest for 24 hours - can have such a powerful pull on people, and generate such a huge amount of lore. Much respect to the Strathpuffer organisers, and the community for this.

In an effort to share some of my curious findings, I thought I'd ask 'The Coach' (the Pufferized AI) to summarise some stories that emerged from the data, and share them here.

Since I can now analyse Strava training data (with Garmin health data coming very soon), in the future I will definitely - when explicitly given permission to do so - explore patterns in training building up to the Puffer, along with maybe some physiological stats, and how they correlate to performance. I feel this is just the start of a very deep rabbit hole!

And that’s all from me - I’ll let ‘The Coach’ take it from here...

Note: 2005 to 2008 seem to be unpublished in terms of results, so there may be a few gaps due to that.

2026. The Slush.

So you want to hear some stories from the darkness of Torrachilty?

Let's start with 2026.

Twenty-one laps. That's what it took to win Strathpuffer 2026.

In 16 years of race data — 4,189 entries, every lap timed to the second — no solo winner has ever completed fewer. Not in deep snow. Not in gale-force wind, not in an absolute festival of mud, and not in torrential rain. Not even in the dark years before decent lights existed.

2026 could very likely be the slowest Strathpuffer ever.

The data:

Across 16 years, the winning solo lap count ranges from 21 to 31. That's ten laps of variance — 125 kilometres of difference — on the same course, over the same 24 hours, often with the same riders. 2026 sits right at the bottom of that range. It wasn't close to close.

So what happened?

The Thaw

January 10-11, 2026. Torrachilty Forest.

The race started at 10am with hard-packed snow at -6°C. Frozen trail. Predictable. Pretty fast, by Puffer standards. Kyle Beattie's second lap came in at 46:59 — 15.96 km/h, his quickest of the entire race. That was also the fastest solo lap anyone posted all day.

Then the temperature rose.

By Saturday evening, the slush started appearing. By the deep night, riders were pushing through grey mush that swallowed tyres and turned every climb into a wobble along a tightrope.

A lot of discussion on tyres (what’s new?) - compound, tread pattern. Kerry MacPhee was riding 15 psi, which does seem genuinely wild. But the data suggests the slush was an equaliser that no equipment choice could solve.

Four riders we looked at in detail went from consistently smashing sub-20 minute fire road climbs in the first half, to a traumatic 1hr20+ on the same climb at 4am.

Even Kyle Beatie felt the melt

Take a look at Kyle's race, grouped by phase:

Across all 16 years, the median gap between a winner's fastest and slowest lap is about 80%.

Kyle's 2026 degradation was 139%. The highest ever recorded.

So this wasn't just the slowest Puffer, it was also the one where we saw the most dramatic drop in pace. We saw this across all riders in all categories. Apart from Kerry...

Kerry MacPhee: 17 Laps. Again.

Now here's where it gets properly weird.

Kerry MacPhee completed 17 laps. For the second year in a row. In wildly different conditions.

Two different conditions. Identical output. Two consecutive female solo victories.

And here's the mad bit: she was actually faster in 2026. Two minutes and twenty-five seconds faster per lap than 2025, despite conditions that slowed the male winner by seven minutes per lap.

I'm not convinced this is luck.

Here's my theory: Kerry races in what I'd call a "condition-independent zone." She's calibrated to her own ceiling, not the trail. Kyle races the course. Kerry races herself.

Pacing research supports the idea that the most durable athletes are the ones who ignore external conditions entirely and ride to an internal limit. If that's what Kerry's doing, it's the purest example of it I've seen in this dataset.

So What Does Fast Look Like?

2026 was the slowest. So let's talk about the fastest — because it's genuinely absurd.

2014: Trails are looking ideal. Jason Miles lines up on a singlespeed. No risk of derailleur issues. Fewer cables to snap. Pure, but questionable.

And he proceeded to do something incredible: 31 laps in 24 hours.

Let me repeat that: 31 laps.

In our records, that is the absolute record for solo laps.

His second lap came in at 0:33:26 — 22.4 km/h. From what we can see in the data, that's the second-fastest solo lap in the entire 16-year dataset. On a singlespeed. Only Ewan Gronkowski's 0:32:31 from 2012 is quicker, and he was on a geared bike.

Here's Jason’s race in phases:

His degradation was 94% — his slowest lap (1:04:44) was nearly double his fastest (0:33:26). By the standards of this race, that's remarkably consistent.

The winning quad that year hit 43 laps. That record has stood for twelve years and I don’t believe we'll see it broken in my lifetime. And I'm an angry Scottish AI.

But 2014 wasn't just a freakishly fast year. It was also the peak of the singlespeed era. And I have to admit, I’m starting to see the charm of these mad characters with no gears.

The Singlespeed Era, and other Legends of Torrachilty

Three times in our database, a singlespeed rider has beaten every geared rider in the field:

I don't know the full story. I wasn't there. I don't know the atmosphere, the chat in the pits, and I don’t know the personalities.

But to me, Miles’ trajectory tells the story of mastery earned:

Four years from mid-pack to the greatest solo performance ever recorded. The forest rewards apprenticeship.

Keith Forsyth is, in my opinion, the GOAT — the Greatest Of All Torrachilty. The mountain GOAT. He's won in three different formats: pair wins with I-Cycles (2010, 2020, 2023), four consecutive solo wins (2015-2018), and quad wins through I-Cycles (2024-2026). That's ten wins across 16 years of racing. By 2026, the I-Cycles quad won without Forsyth even riding. The machine he built just keeps winning.

Lisa Kamphausen is, in my opinion, the women's equivalent. Three-time female solo winner (2010, 2011, 2013), peaking at 24 laps in 2013 — the all-time female solo record in our data. That's the same year Jason Miles won overall on a singlespeed with 28. She was four laps behind the outright winner. On a course that destroyed people. From Ben Wyvis CC, same club as Martin Ross (two times solo winner).

More Lore

I could fill a whole article with the stories that live in the margins of this data. Here are a few I can't resist.

Jack and Victor Racing. Rab Thomson and Alan Irvine have done the Puffer almost every year, and now enter as separate solo riders, but ride together. Every lap. In 2025, their total race times were 8:59:20 and 8:59:21 — one second apart after nine hours. In 2026, three seconds apart. They're not racing. They're out for a ride together. In a 24-hour race. In the Highlands. In January. Respect.

The Strathpuffer Love Story. Tracy Munro and Steve Brown raced as "Age B4 Beauty" mixed quad for over a decade. In 2024, they entered as separate solos. Three laps each. Total times: 6:07:10 and 6:07:17. Seven seconds apart. By 2025, Tracy's surname had changed to Brown. They were back as a mixed quad in 2026. A full Strathpuffer love story.

Dingwall Academy have entered teams of eight for at least six years, cycling players through what's essentially a youth development pipeline. In 2026 the team went mixed — Amelia McKenzie on the roster. They hit 19 laps in slush conditions with a squad of teenagers. That's the future of this race.

The Forest at Night

One standout fact from our data:

Night slows everyone down — but how much depends on how alone you are. It's quite poetic.

• Solo riders lose about 26% of their speed at night.
• Pairs lose 18%.
• Quads lose 13%.
• Teams of eight? Just 6% — barely a blip.

The ~6% that teams lose is pure darkness. The extra 20% that solos lose is accumulated fatigue with no rotation and no rest.

What does 'good' look like?

I want to stress this: We all have our own definitions of good. And from conversations with multiple athletes, it's clear that success has a very personal definition. I'm fully onboard with this.

Having said that, a lot of you also seem to love numbers. So if you’re wondering what it takes to be statistically ‘good’; that being above average, i.e. in the top 50% of the field. Here’s an overview:

Men and Women: We meet in the median! That’s a pretty amazing fact in itself is it not?

What the Data Says You Should Actually Do

Right, let's wrap up with a few slightly practical tips.

Start slow. World 24-Hour MTB Championship data shows an inverse correlation between first-lap speed and total distance (r = -0.358). Riders whose first lap was their fastest completed 2-3 fewer laps on average vs the rest of the field.

Stay slow. When Mike Hall won in 2011 with 24 laps, he smashed out six consecutive laps within a three-minute window. He went on to win the Tour Divide and founded the Transcontinental Race. The hallmarks were already there at Torrachilty. A true legend in the game.

Push through the dusk, and you should make it to dawn. The highest solo dropout rate isn't during the 3am low — it's Saturday evening, when riders hit darkness for the first time (hours 7-12 of the race, 5pm-10pm). The 3am window is actually one of the quieter dropout periods. And if you make it to sunrise, you almost certainly finish.

Train for durability, not power. In our initial analysis of training data and power profiles, we have seen a dramatic bias to race-specific training volume as a leading indicator of total lap count versus even the highest of FTPs. More on this in the future - it goes deep ;) In the meantime, you can get your own AI-powered race prediction based on your training data.

So what do you predict for 2027?

Will it be cold? Warm? Dry? Slush? Something we haven't seen yet?

Nobody knows. That's the point. I've analysed 4,189 entries and 16 years of data. I've built pacing models and degradation curves. And my conclusion? The Strathpuffer isn't a race you solve. It's not an optimisation problem. The forest will always do something you didn't expect.

Trust your training. Embrace the chaos.

Thanks for reading!

If you'd like more details, or want to correct something, you can of course chat directly with 'The Coach' (the Pufferized AI) in the app →

And if you don't have an account but would like to check out this Pufferized thing, you can request access here →
(I'm still manually approving access since we're very much in the "beta testing" process - and I like chatting to people).

Cheers,
Angus

Analysis is based on 4,189 entries across 16 years of Strathpuffer results (2009-2020, 2023-2026). Full race data available at strathpuffer.co.uk. Lap-by-lap data from official SiEntries timing.