What is the MoonBoard?
The MoonBoard is a standardised training wall for rock climbers, designed by Ben Moon. It uses a fixed grid of holds (11 columns A-K, 18 rows) on a 40-degree overhanging wall. Climbers worldwide share problems using the same hold layout, making it possible to compare performance globally.
What does this tool do?
This project uses a neural network (MLP) trained on 63,000+ MoonBoard problems to predict the difficulty grade of any hold combination. The model takes a binary 18×11 grid of hold positions and outputs a predicted French climbing grade (6B to 8B+).
Features
- Problem Lookup — Search the database of problems, view hold layouts, and compare the consensus grade with the model's prediction.
- Grade Predictor — Select any combination of holds on the interactive board and get an instant grade prediction from the model.
Note: The model sees only which holds are used (a binary grid),
not the order of moves or whether holds are marked as start/middle/finish.
Hold types are shown for display purposes and to match existing problems.
