Beyond Research

Cycling, the outdoors, and the occasional data question.

Outside my academic work, my main hobby is cycling, and now and then it overlaps with the day job.

Cycling

I ride bikes, both outdoors and indoors on Zwift, and I try to put in 8 to 12 hours a week, though the honest average is closer to 8 or 9; a good week hits 12, but life tends to get in the way. Most of it is around Fribourg, where I live: into the hills and mountains when I want to climb, and out toward the Murtensee (Lake Murten) when I want flatter, faster roads. Off the bike, I also hike, nothing too extreme, and go for the occasional run. I log it all on Strava, if you want to follow along.

On the carbon bike near Schwarzenburg, the Préalpes behind. Spring 2026.
Climbing above Lake Geneva, in the Lavaux vineyards (Puidoux). Spring 2026.
Hiking in the Alps. Summer 2025.
Above Schwarzsee, in the Fribourg Préalpes. Spring 2024.

The intersection between statistics and my hobby

Recording every ride for years leaves me with a lot of data, so I analyzed it the way I would any other dataset. The first thing I tested: whether the expensive carbon wheels I bought are actually faster, or mostly marketing.

On 21 April 2026 I replaced my road bike’s aluminium wheels with deep-section carbon ones and left everything else unchanged. Adjusting for power and terrain, the carbon setup is the fastest of my three bikes, about +1.2 km/h quicker than the same road bike on its old alloy wheels.

Adjusted average speed by setup, holding power and terrain at their means. The carbon road bike is the fastest of the three, about +1.2 km/h ahead of the same bike on alloy wheels. Bars are 95% confidence intervals.

That could be the wheels, but it could also just be that I was fitter and the weather was better. To separate the two, I use my gravel bike, ridden on the same roads in the same weeks and never upgraded, as a control: it should not speed up unless the season, rather than the wheels, is doing the work. In a difference-in-differences after the install, the road bike gains +1.7 km/h and the gravel bike does not.

Road bike versus the gravel-bike placebo (difference-in-differences, season and year fixed effects, at equal power and terrain). After the install the road bike gains +1.7 km/h while the never-upgraded gravel bike does not (-1.2). Bars are 95% confidence intervals.

Comparing like season with like season, though, most of that gap goes away, so I wouldn’t read too much into six weeks of data.

I wrote it up as a short working paper, with real tables, a placebo, and a conflict-of-interest statement (I own the wheels):

📄 Working paper (PDF)

Half-serious. The regressions are real.