Overfitting and Leaderboard shakeup

Hi guys,

Do you think there is going to be a lot of shakeup in the leaderboard in this competition? I feel like it is very easy to overfit and have a submission that works really well on the small hidden sample but may fail significantly on the full dataset. I am also seeing better local CV scores so it makes me wonder.

1 Like

According to the variance I’ve observed between CV and public LB on my model, I think shake up could happen but winners should remain in the top 11 we’ve currently (i.e. between 0.85 and 0.87).

Wait and see. Good luck to everyone! It was a very nice competition.