**Scoring and test dataset details**: For both tracks, 20% of the Sygen data will be held back as test data and used in the evaluation process. The Kaggle public leaderboard is computed on a subset of the test data for the functional prediction task, so while it will give you a sense of your relative performance it may not exactly match the final evaluation. This prevents overfitting through submissions.

**Functional Prediction**: For prediction of walking performance, which uses an ordinal scale, the Spearman’s rank correlation coefficient will be computed between the predicted and actual scores. This correlation will be the score for the submission.

We provide a template for the submission file in the Data section.