Rataplat is a climbing recommender for boulderers, built around a simple problem: at a huge area like Fontainebleau there are thousands of problems, and finding the ones you will actually enjoy is mostly luck.
How it works
- Every climb is tagged by style: crimpy, slopey, compression, prow, and so on.
- Climbers log what they send and what they think of it, and the tags they add are crowdsourced back into the database.
- The model learns a climber’s strengths and tastes from their log and suggests climbs at an area that match.
Status
The first version of the recommender was a content-based model in R. I am rebuilding it in Python and releasing it as an Android app. Graphs of a climber’s strengths and a feed of what similar climbers are sending are planned for later versions.