It is the first time for me to participated in Kaggle competition. After months of effort, I got LB=0.492 in Dec., located in a position of over 300 on leaderboard. There're 300+ 0.494s ranked before me. I guessed these 0.494s were all copied from the shared code. When I finally got LB=0.495 on Dec. 10, I was ranked 27th which made me so excited. I overcame the obstacle of 0.494s. But today, I logined and found I fell to 350th again. There're 300+ 0.497s once again! Only because the top 1 shared code updated and got 0.497.
\nIs it fair? We had put in months of hard work to improve by one thousandth, then were defeated by hundreds of copies in an instant. Are Kaggle's rules meant to encourage this situation?
", "messages": [ { "id": "3072185", "postDate": "12/14/2024 20:29:38", "content": "It is the first time for me to participated in Kaggle competition. After months of effort, I got LB=0.492 in Dec., located in a position of over 300 on leaderboard. There're 300+ 0.494s ranked before me. I guessed these 0.494s were all copied from the shared code. When I finally got LB=0.495 on Dec. 10, I was ranked 27th which made me so excited. I overcame the obstacle of 0.494s. But today, I logined and found I fell to 350th again. There're 300+ 0.497s once again! Only because the top 1 shared code updated and got 0.497.
\nIs it fair? We had put in months of hard work to improve by one thousandth, then were defeated by hundreds of copies in an instant. Are Kaggle's rules meant to encourage this situation?
", "rawMarkdown": "It is the first time for me to participated in Kaggle competition. After months of effort, I got LB=0.492 in Dec., located in a position of over 300 on leaderboard. There're 300+ 0.494s ranked before me. I guessed these 0.494s were all copied from the shared code. When I finally got LB=0.495 on Dec. 10, I was ranked 27th which made me so excited. I overcame the obstacle of 0.494s. But today, I logined and found I fell to 350th again. There're 300+ 0.497s once again! Only because the top 1 shared code updated and got 0.497. \n\nIs it fair? We had put in months of hard work to improve by one thousandth, then were defeated by hundreds of copies in an instant. Are Kaggle's rules meant to encourage this situation?", "votes": null }, { "id": "3072200", "postDate": "12/14/2024 21:00:29", "content": "Well, this is why sharing in the last week is forbidden.
\nNow I'm gonna try to improve this shared solution…I have no idea what is going on in this competition, but I also want to participate in this fun and cute lottery competition! (at least this is what it looks like?)
\n\nWe had put in months of hard work to improve by one thousandth,
\n
This is not that kind of competition where you should steadily climb to LB position and care about it so much. What you really need to think about is the robustness of the solution. A balanced trade-off is all you need, because you can't go through chaos if you are too conservative.
", "rawMarkdown": ">We had put in months of hard work to improve by one thousandth,\n\nThis is not that kind of competition where you should steadily climb to LB position and care about it so much. What you really need to think about is the robustness of the solution. A balanced trade-off is all you need, because you can't go through chaos if you are too conservative.", "votes": null }, { "id": "3072245", "postDate": "12/14/2024 22:57:31", "content": "Why not add something you found to this public code?
", "rawMarkdown": "Why not add something you found to this public code?", "votes": null }, { "id": "3072371", "postDate": "12/15/2024 05:02:27", "content": "This is the fastest way to get notebook(discussion) GM. Sharing ensemble notebooks with a high score but full of unreasonable code or writing down thousands of waffle, you could be entitled as a \"GM\".
\nYou need to distinguish which notebooks are valuable and helpful for private leaderboard.
Usually the public notebook with high public LB doesnt perform well eventually, so take it easy. But with that being said, this is one of the most fun competitions I've ever encountered… really looking forward to seeing the outcome
", "rawMarkdown": "Usually the public notebook with high public LB doesnt perform well eventually, so take it easy. But with that being said, this is one of the most fun competitions I've ever encountered... really looking forward to seeing the outcome", "votes": null }, { "id": "3072407", "postDate": "12/15/2024 06:17:36", "content": "No problem, what truly matters is not your ranking on the LB, but your ranking on the PB and your personal growth.
", "rawMarkdown": "No problem, what truly matters is not your ranking on the LB, but your ranking on the PB and your personal growth.", "votes": null }, { "id": "3072486", "postDate": "12/15/2024 09:30:08", "content": "There are two very decent notebooks that is shared in this competition thanks to @ambrosm and @antoninadolgorukova. Normally, you don't see that often in lottery competitions.
\nThey are truly great. And I have learnt a lot from AmbrosM's notebooks in many competitions.
", "rawMarkdown": "They are truly great. And I have learnt a lot from AmbrosM's notebooks in many competitions.", "votes": null }, { "id": "3073096", "postDate": "12/16/2024 03:56:15", "content": "being good is not enough. you have to be better than the rest in a competition.
\nIn today's LLM/chatgpt age, anyone (skilled or less skilled, spent much time or less time) who knows how to use AI as a tool will have the same score.
await shake up
", "rawMarkdown": "await shake up", "votes": null }, { "id": "3076532", "postDate": "12/20/2024 03:14:26", "content": "Hey, your hard work prevailed over our copy pasta shenanigans.. congratulations for medal win
", "rawMarkdown": "Hey, your hard work prevailed over our copy pasta shenanigans.. congratulations for medal win", "votes": null }, { "id": "3076567", "postDate": "12/20/2024 04:36:43", "content": "The open-source 0.497 code didn't score well on the private leaderboard. Instead, those who last logged in over a month ago and submitted only once or twice ended up winning gold medals. I think the randomness of such competitions takes away much of the fun, though you can still learn a lot from them. All in all, it's probably best to avoid these kinds of \"lottery-style competitions\" in the future.
", "rawMarkdown": "The open-source 0.497 code didn't score well on the private leaderboard. Instead, those who last logged in over a month ago and submitted only once or twice ended up winning gold medals. I think the randomness of such competitions takes away much of the fun, though you can still learn a lot from them. All in all, it's probably best to avoid these kinds of \"lottery-style competitions\" in the future.", "votes": null } ], "comments": [ { "id": 3072200, "author_name": "shlomoron", "author_url": "", "post_date": "12/14/2024 21:00:29", "content": "Well, this is why sharing in the last week is forbidden.
\nNow I'm gonna try to improve this shared solution…I have no idea what is going on in this competition, but I also want to participate in this fun and cute lottery competition! (at least this is what it looks like?)
\n\nWe had put in months of hard work to improve by one thousandth,
\n
This is not that kind of competition where you should steadily climb to LB position and care about it so much. What you really need to think about is the robustness of the solution. A balanced trade-off is all you need, because you can't go through chaos if you are too conservative.
", "votes": null, "replies": [] }, { "id": 3072245, "author_name": "aisysxd", "author_url": "", "post_date": "12/14/2024 22:57:31", "content": "Why not add something you found to this public code?
", "votes": null, "replies": [] }, { "id": 3072371, "author_name": "takanashihumbert", "author_url": "", "post_date": "12/15/2024 05:02:27", "content": "This is the fastest way to get notebook(discussion) GM. Sharing ensemble notebooks with a high score but full of unreasonable code or writing down thousands of waffle, you could be entitled as a \"GM\".
\nYou need to distinguish which notebooks are valuable and helpful for private leaderboard.
There are two very decent notebooks that is shared in this competition thanks to @ambrosm and @antoninadolgorukova. Normally, you don't see that often in lottery competitions.
\nThey are truly great. And I have learnt a lot from AmbrosM's notebooks in many competitions.
", "votes": null, "replies": [] } ] } ] }, { "id": 3072375, "author_name": "yannan90", "author_url": "", "post_date": "12/15/2024 05:06:14", "content": "Usually the public notebook with high public LB doesnt perform well eventually, so take it easy. But with that being said, this is one of the most fun competitions I've ever encountered… really looking forward to seeing the outcome
", "votes": null, "replies": [] }, { "id": 3072407, "author_name": "yosynory", "author_url": "", "post_date": "12/15/2024 06:17:36", "content": "No problem, what truly matters is not your ranking on the LB, but your ranking on the PB and your personal growth.
", "votes": null, "replies": [] }, { "id": 3073096, "author_name": "hengck23", "author_url": "", "post_date": "12/16/2024 03:56:15", "content": "being good is not enough. you have to be better than the rest in a competition.
\nIn today's LLM/chatgpt age, anyone (skilled or less skilled, spent much time or less time) who knows how to use AI as a tool will have the same score.
await shake up
", "votes": null, "replies": [] }, { "id": 3076532, "author_name": "jagatkiran", "author_url": "", "post_date": "12/20/2024 03:14:26", "content": "Hey, your hard work prevailed over our copy pasta shenanigans.. congratulations for medal win
", "votes": null, "replies": [] }, { "id": 3076567, "author_name": "kitsuha", "author_url": "", "post_date": "12/20/2024 04:36:43", "content": "The open-source 0.497 code didn't score well on the private leaderboard. Instead, those who last logged in over a month ago and submitted only once or twice ended up winning gold medals. I think the randomness of such competitions takes away much of the fun, though you can still learn a lot from them. All in all, it's probably best to avoid these kinds of \"lottery-style competitions\" in the future.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3072185": "It is the first time for me to participated in Kaggle competition. After months of effort, I got LB=0.492 in Dec., located in a position of over 300 on leaderboard. There're 300+ 0.494s ranked before me. I guessed these 0.494s were all copied from the shared code. When I finally got LB=0.495 on Dec. 10, I was ranked 27th which made me so excited. I overcame the obstacle of 0.494s. But today, I logined and found I fell to 350th again. There're 300+ 0.497s once again! Only because the top 1 shared code updated and got 0.497. \n\nIs it fair? We had put in months of hard work to improve by one thousandth, then were defeated by hundreds of copies in an instant. Are Kaggle's rules meant to encourage this situation?", "3072200": "Well, this is why sharing in the last week is forbidden. \nNow I'm gonna try to improve this shared solution...I have no idea what is going on in this competition, but I also want to participate in this fun and cute lottery competition! (at least this is what it looks like?)", "3072212": ">We had put in months of hard work to improve by one thousandth,\n\nThis is not that kind of competition where you should steadily climb to LB position and care about it so much. What you really need to think about is the robustness of the solution. A balanced trade-off is all you need, because you can't go through chaos if you are too conservative.", "3072245": "Why not add something you found to this public code?", "3072371": "This is the fastest way to get notebook(discussion) GM. Sharing ensemble notebooks with a high score but full of unreasonable code or writing down thousands of waffle, you could be entitled as a \"GM\".\nYou need to distinguish which notebooks are valuable and helpful for private leaderboard.", "3072375": "Usually the public notebook with high public LB doesnt perform well eventually, so take it easy. But with that being said, this is one of the most fun competitions I've ever encountered... really looking forward to seeing the outcome", "3072407": "No problem, what truly matters is not your ranking on the LB, but your ranking on the PB and your personal growth.", "3072486": "There are two very decent notebooks that is shared in this competition thanks to @ambrosm and @antoninadolgorukova. Normally, you don't see that often in lottery competitions. \n* https://www.kaggle.com/code/ambrosm/piu-eda-which-makes-sense\n* https://www.kaggle.com/code/antoninadolgorukova/cmi-piu-features-eda", "3072498": "They are truly great. And I have learnt a lot from AmbrosM's notebooks in many competitions.", "3073096": "being good is not enough. you have to be better than the rest in a competition.\nIn today's LLM/chatgpt age, anyone (skilled or less skilled, spent much time or less time) who knows how to use AI as a tool will have the same score.", "3075358": "await shake up", "3076532": "Hey, your hard work prevailed over our copy pasta shenanigans.. congratulations for medal win", "3076567": "The open-source 0.497 code didn't score well on the private leaderboard. Instead, those who last logged in over a month ago and submitted only once or twice ended up winning gold medals. I think the randomness of such competitions takes away much of the fun, though you can still learn a lot from them. All in all, it's probably best to avoid these kinds of \"lottery-style competitions\" in the future." }, "source": "meta" }