what type of data are you looking for? power, time, distance etc.? drag coefficient? what are you trying to predictively model?

I'm not aware of any public datasets but the guys at golden cheetah might be able to provide you a scrubbed csv file or python/R dataframe to get you started.

---

the world's still turning? >>>>>>> the world's still turning

no particular objective, I'm just looking to get a model working on HuggingFace and it may as well be something I'm interested in :-)

The Golden Cheetah folks are a good idea!

---

" I take my gear out of my car and put my bike together. Tourists and locals are watching from sidewalk cafes. Non-racers. The emptiness of of their lives shocks me. "
(opening lines from Tim Krabbe's The Rider , 1978

There are none, but some things are scrapable - race results for example.

---

Next races on the schedule: none at the moment

Strava has a decent API. This site is easily scrapeable.

Entalpi put out a public dataset from blummenfelt and idens kona run data not long ago, not big enough to do true ML with but still fun.
https://github.com/entalpi-no/kona-2022

While I would not necessarily consider myself an expert, I am doing my PhD thesis using machine learning methods. Additionally, as any good ST participant, I am a fan of the sport we call triathlon. The biggest issue with applying ML algorithms is that you need a lot of data, which in this case may be hard to come by. I analyzed a dataset relating the number of deaths in our sport and broke it down to when they occurred, however that was basic statistical analyses and not using ML as it was only a few hundred points I believe. If interested I can look up said database and you can do some lite ML analytics I suppose.

Any other Qs feel free to reach out!

---

Today I do what others won't, so tomorrow I can do what others can't.

I recently got one of my Post-Doc researchers to start using ML on very large data sets (economic and engineering data). In the past, I’ve used AI in some published papers though I am typically a traditional time series analysis/econometrics guy. I’d suggest starting out with something like LASSO - pretty easy to implement and then you can ramp things up from there.

---

"The more you suffer, the more it shows you really care.”

Hi. I'm involved in a related start-up that's trucking along and so read this post with interest. Our team houses experts that are some of the best in the world in this field (you can research this on our staff page). If you or anyone reading this thread is interested in a project, please get in touch with me. paul at athletica dot ai

---

https://athletica.ai/
https://hiitscience.com/