Hi, Organizers,   I have read about that 'one missing gene'. My teammate spent tons of time to figure out this in order to run through this little problem, and until now, I am not entirely convinced what we are running through is correct because of this discrepancy on your side. Due to this gene, we didn't even have time to improve our model in the end. I think it shall be extremely unfair to us if deadline is extended for this problem, as our time was spent here.   However, it will be extremely unfair to other teams that due to this one gene their submission isn't even scored. They didn't do anything wrong.   This is a great opportunity that your docker system would shine: You have the the submission IDs, from us and from others. Now, all you need to do, is fix the data together with evaluation in two different versions, 10005 or 10006, and run the submission against it. There is no reason that a submission doesn't run through any of them (that case, the team has other bugs that should have been fixed long ago, they deserve to be removed). If only one runs through, you take that score. If both runs through, you take the max (because the other one could be half-random due to one line shifting).   This should solve all concerns and justify the docker system.