{ "id": 509133, "title": "True winners are #2, #4, #7 and #9", "url": "/competitions/home-credit-credit-risk-model-stability/discussion/509133", "author_name": "", "post_date": "2024-06-01T10:46:19.442466900Z", "votes": -11, "comment_count": 3, "views": 0, "content": "

Submitting a solution that ranks around 1200 in the public leaderboard takes a lot of conviction that what you do is right!

\n

Congratulations to @d84136957, @flashba, @tttaro, @ningzhao28, @xinyueou, @yangxintong1010, @qscrow, @xdxfory, @idiotcoke and @scoooot. It would be great if you could publish your solutions.

", "messages": [ { "id": "2849022", "postDate": "06/01/2024 10:46:19", "content": "

Submitting a solution that ranks around 1200 in the public leaderboard takes a lot of conviction that what you do is right!

\n

Congratulations to @d84136957, @flashba, @tttaro, @ningzhao28, @xinyueou, @yangxintong1010, @qscrow, @xdxfory, @idiotcoke and @scoooot. It would be great if you could publish your solutions.

", "rawMarkdown": "Submitting a solution that ranks around 1200 in the public leaderboard takes a lot of conviction that what you do is right!\n\nCongratulations to @d84136957, @flashba, @tttaro, @ningzhao28, @xinyueou, @yangxintong1010, @qscrow, @xdxfory, @idiotcoke and @scoooot. It would be great if you could publish your solutions.", "votes": null }, { "id": "2849392", "postDate": "06/01/2024 15:05:35", "content": "

Only because they jumped ranks doesn't mean that it's better. Not jumping ranks is even more likely to be a stable model.

", "rawMarkdown": "Only because they jumped ranks doesn't mean that it's better. Not jumping ranks is even more likely to be a stable model.", "votes": null }, { "id": "2849833", "postDate": "06/01/2024 18:58:49", "content": "

I am not saying that they are better. I say you need to be truly believing in your work to stick to it even if it does not show performance in public leaderboard.

\n

Lot of guys here experienced best private performance from unsubmitted models. So my point is that public leaderboard is not the best indicator of success. You need to know what you do and why you do it.

", "rawMarkdown": "I am not saying that they are better. I say you need to be truly believing in your work to stick to it even if it does not show performance in public leaderboard.\n\nLot of guys here experienced best private performance from unsubmitted models. So my point is that public leaderboard is not the best indicator of success. You need to know what you do and why you do it.", "votes": null }, { "id": "2852608", "postDate": "06/03/2024 11:24:24", "content": "

Once again, truly believing and to know what you do, it can be also coincidence that people just submit whatever ranked in public LB and then jump up by luck. This isn't any indicator and therefore not a valid argument to degrade the winners and call them \"not true\" winners and list other \"true\" winners. That simply a bold and rude statement.

", "rawMarkdown": "Once again, truly believing and to know what you do, it can be also coincidence that people just submit whatever ranked in public LB and then jump up by luck. This isn't any indicator and therefore not a valid argument to degrade the winners and call them \"not true\" winners and list other \"true\" winners. That simply a bold and rude statement.", "votes": null } ], "comments": [ { "id": 2849392, "author_name": "olafgundersen", "author_url": "", "post_date": "06/01/2024 15:05:35", "content": "

Only because they jumped ranks doesn't mean that it's better. Not jumping ranks is even more likely to be a stable model.

", "votes": null, "replies": [ { "id": 2849833, "author_name": "svendaj", "author_url": "", "post_date": "06/01/2024 18:58:49", "content": "

I am not saying that they are better. I say you need to be truly believing in your work to stick to it even if it does not show performance in public leaderboard.

\n

Lot of guys here experienced best private performance from unsubmitted models. So my point is that public leaderboard is not the best indicator of success. You need to know what you do and why you do it.

", "votes": null, "replies": [ { "id": 2852608, "author_name": "olafgundersen", "author_url": "", "post_date": "06/03/2024 11:24:24", "content": "

Once again, truly believing and to know what you do, it can be also coincidence that people just submit whatever ranked in public LB and then jump up by luck. This isn't any indicator and therefore not a valid argument to degrade the winners and call them \"not true\" winners and list other \"true\" winners. That simply a bold and rude statement.

", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2849022": "Submitting a solution that ranks around 1200 in the public leaderboard takes a lot of conviction that what you do is right!\n\nCongratulations to @d84136957, @flashba, @tttaro, @ningzhao28, @xinyueou, @yangxintong1010, @qscrow, @xdxfory, @idiotcoke and @scoooot. It would be great if you could publish your solutions.", "2849392": "Only because they jumped ranks doesn't mean that it's better. Not jumping ranks is even more likely to be a stable model.", "2849833": "I am not saying that they are better. I say you need to be truly believing in your work to stick to it even if it does not show performance in public leaderboard.\n\nLot of guys here experienced best private performance from unsubmitted models. So my point is that public leaderboard is not the best indicator of success. You need to know what you do and why you do it.", "2852608": "Once again, truly believing and to know what you do, it can be also coincidence that people just submit whatever ranked in public LB and then jump up by luck. This isn't any indicator and therefore not a valid argument to degrade the winners and call them \"not true\" winners and list other \"true\" winners. That simply a bold and rude statement." }, "source": "meta" }