perhaps the date is not checked?

This competition uses an external metric we set up outside the normal Zindi metric system, and it looks like the scoring function assumes the dataframe is ordered by date as in the submission file. Normally we can ignore row order because we're matching rows on an ID, here the idea is to look for the last specified location before a crash happens. I'll double-check and then get the team to put out a notice. My suggestion for now is to follow the ordering from the sample submission.