{ "id": 287217, "title": "Will the public leaderboard remain unknown forever?", "url": "/competitions/g-research-crypto-forecasting/discussion/287217", "author_name": "", "post_date": "2021-11-12T18:17:21.308079Z", "votes": 1, "comment_count": 1, "views": 0, "content": "

Will the public leaderboard be unknown for the whole timeline of this competition? or will we be able to see it before the final submission deadline?

\n

I'm asking because if one does not know what is his score compared to other participants, he won't know if he has to improve his solution.

", "messages": [ { "id": "1580416", "postDate": "11/12/2021 18:17:21", "content": "

Will the public leaderboard be unknown for the whole timeline of this competition? or will we be able to see it before the final submission deadline?

\n

I'm asking because if one does not know what is his score compared to other participants, he won't know if he has to improve his solution.

", "rawMarkdown": "Will the public leaderboard be unknown for the whole timeline of this competition? or will we be able to see it before the final submission deadline?\n\nI'm asking because if one does not know what is his score compared to other participants, he won't know if he has to improve his solution.", "votes": null }, { "id": "1580791", "postDate": "11/13/2021 05:43:13", "content": "

The public leaderboard is visible to everyone for the entirety of the competition see this link. They cautioned against using the public leaderboard to see how you are going as the test data is essentially the same as the training data. This discussion post by the competition host says:

\n
\n

It is worth reiterating that - ‘This public leaderboard uses publicly available data. Ranks here are not meaningful. Furthermore, the public leaderboard targets are publicly available and are provided as part of the competition dataset. Expect to see many people submitting perfect submissions for fun.’

\n
\n

Only when the competition is finalised will the actual test data be used to run our models against. Having a high rank in the public leaderboard is not always an indicator of success in the private leaderboard and can indicate overfitting of your model against the training/test data which will not provide a good generalised model.

\n

Good luck with the competition.

", "rawMarkdown": "The public leaderboard is visible to everyone for the entirety of the competition see this [link](https://www.kaggle.com/c/g-research-crypto-forecasting/leaderboard). They cautioned against using the public leaderboard to see how you are going as the test data is essentially the same as the training data. This [discussion post](https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/284848) by the competition host says:\n> It is worth reiterating that - ‘This public leaderboard uses publicly available data. Ranks here are not meaningful. Furthermore, the public leaderboard targets are publicly available and are provided as part of the competition dataset. Expect to see many people submitting perfect submissions for fun.’ \n\nOnly when the competition is finalised will the actual test data be used to run our models against. Having a high rank in the public leaderboard is not always an indicator of success in the private leaderboard and can indicate overfitting of your model against the training/test data which will not provide a good generalised model. \n\nGood luck with the competition.", "votes": null } ], "comments": [ { "id": 1580791, "author_name": "allanbond", "author_url": "", "post_date": "11/13/2021 05:43:13", "content": "

The public leaderboard is visible to everyone for the entirety of the competition see this link. They cautioned against using the public leaderboard to see how you are going as the test data is essentially the same as the training data. This discussion post by the competition host says:

\n
\n

It is worth reiterating that - ‘This public leaderboard uses publicly available data. Ranks here are not meaningful. Furthermore, the public leaderboard targets are publicly available and are provided as part of the competition dataset. Expect to see many people submitting perfect submissions for fun.’

\n
\n

Only when the competition is finalised will the actual test data be used to run our models against. Having a high rank in the public leaderboard is not always an indicator of success in the private leaderboard and can indicate overfitting of your model against the training/test data which will not provide a good generalised model.

\n

Good luck with the competition.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1580416": "Will the public leaderboard be unknown for the whole timeline of this competition? or will we be able to see it before the final submission deadline?\n\nI'm asking because if one does not know what is his score compared to other participants, he won't know if he has to improve his solution.", "1580791": "The public leaderboard is visible to everyone for the entirety of the competition see this [link](https://www.kaggle.com/c/g-research-crypto-forecasting/leaderboard). They cautioned against using the public leaderboard to see how you are going as the test data is essentially the same as the training data. This [discussion post](https://www.kaggle.com/c/g-research-crypto-forecasting/discussion/284848) by the competition host says:\n> It is worth reiterating that - ‘This public leaderboard uses publicly available data. Ranks here are not meaningful. Furthermore, the public leaderboard targets are publicly available and are provided as part of the competition dataset. Expect to see many people submitting perfect submissions for fun.’ \n\nOnly when the competition is finalised will the actual test data be used to run our models against. Having a high rank in the public leaderboard is not always an indicator of success in the private leaderboard and can indicate overfitting of your model against the training/test data which will not provide a good generalised model. \n\nGood luck with the competition." }, "source": "meta" }