This repository contains my submission to the IEEE BigData 2024 Cup: Predicting Chess Puzzle Difficulty, where I placed 4th.
The solution is based on a learning-to-rank approach.
glicko_cnn.ipynb
: Notebook used to train the modelglicko_inference.ipynb
: Notebook for generating predictions on test data
The core idea is to frame puzzle difficulty prediction as a pairwise ranking problem, to predict the relative difficulty between puzzle pairs rather than absolute scores.
You can find a detailed description of the approach in the following resources:
- π Paper on IEEE Xplore
- π Paper (PDF)
- π€ Presentation slides