![\"\"](\"https://i.imgur.com/1Zn9ZgB.png\")
After seeing the published work with a 0.8 score today, it became very unpleasant for me, as for many others, I think.
\nThis behavior ruins the whole point of the competition for both participants and organizers, where only the top ~150 have a chance to prove themselves. I spent a lot of time training models, experimenting with features and studying trees in details. And how terrible it is to realize that after all this, someone is breathing down your back, who decided to amuse his ego, taking with him 100+ participants who, perhaps, didn't even do anything, downloaded the file - uploaded it - went to drink tea waiting for a new file.
\nI am eternally grateful to the organizers for this competition, which allowed me to discover so many new, useful and interesting things. Nevertheless, there remains a strong sediment from the behavior of some participants.
\n☹️
", "messages": [ { "id": 1900158, "postDate": "2022-08-15T18:40:36.900Z", "content": "After seeing the published work with a 0.8 score today, it became very unpleasant for me, as for many others, I think.
\nThis behavior ruins the whole point of the competition for both participants and organizers, where only the top ~150 have a chance to prove themselves. I spent a lot of time training models, experimenting with features and studying trees in details. And how terrible it is to realize that after all this, someone is breathing down your back, who decided to amuse his ego, taking with him 100+ participants who, perhaps, didn't even do anything, downloaded the file - uploaded it - went to drink tea waiting for a new file.
\nI am eternally grateful to the organizers for this competition, which allowed me to discover so many new, useful and interesting things. Nevertheless, there remains a strong sediment from the behavior of some participants.
\n☹️
", "rawMarkdown": "After seeing the published work with a 0.8 score today, it became very unpleasant for me, as for many others, I think. \n\nThis behavior ruins the whole point of the competition for both participants and organizers, where only the top ~150 have a chance to prove themselves. I spent a lot of time training models, experimenting with features and studying trees in details. And how terrible it is to realize that after all this, someone is breathing down your back, who decided to amuse his ego, taking with him 100+ participants who, perhaps, didn't even do anything, downloaded the file - uploaded it - went to drink tea waiting for a new file.\n\nI am eternally grateful to the organizers for this competition, which allowed me to discover so many new, useful and interesting things. Nevertheless, there remains a strong sediment from the behavior of some participants.\n\n☹️", "votes": 18 }, { "id": 1900271, "postDate": "2022-08-15T21:14:30.130Z", "content": "I totally agree, this ruins the competitive spirit of many competitions for ~90% of participants. However, I am still unsure how to deal with this since especially beginners can learn a lot from public notebooks. Maybe the block for publishing top-scoring notebooks should come in earlier.
", "rawMarkdown": "I totally agree, this ruins the competitive spirit of many competitions for ~90% of participants. However, I am still unsure how to deal with this since especially beginners can learn a lot from public notebooks. Maybe the block for publishing top-scoring notebooks should come in earlier. ", "votes": 7 }, { "id": 1900164, "postDate": "2022-08-15T18:46:10.853Z", "content": "I definitely agree. Having read about what happened here, I am appalled at this behavior. While this may be a competition, it is meant for people to test their skills, not engage in what is borderline sabotage. One of my favorite part of Kaggle is how smart, yet incredibly helpful the community is, especially towards someone like me who doesn't have much experience.
", "rawMarkdown": "I definitely agree. Having read about what happened here, I am appalled at this behavior. While this may be a competition, it is meant for people to test their skills, not engage in what is borderline sabotage. One of my favorite part of Kaggle is how smart, yet incredibly helpful the community is, especially towards someone like me who doesn't have much experience. ", "votes": 1, "replies": [ { "id": 1900170, "postDate": "2022-08-15T18:49:35.667Z", "content": "The question is, why would you use a public submission file if you are here to test your skills? Should you not be worrying about your own work? Trust your CV not the public LB?
\nI think you and OP seem to not be aware only the private LB is what matters..?
\nHere is a little trick to help you out if you deeply care about overfitting to the public LB:
\n
@julianmukaj You probably don't understand at all why everyone is so angry at you… competition as such is an excellent motivator for mastering new knowledge. And people like you discourage this motivation from less powerful participants.
", "rawMarkdown": "@julianmukaj You probably don't understand at all why everyone is so angry at you... competition as such is an excellent motivator for mastering new knowledge. And people like you discourage this motivation from less powerful participants.", "votes": 6 }, { "id": 1900194, "postDate": "2022-08-15T19:08:42.170Z", "content": "I have been on the other side of this in other competitions too, a public LB being wiped should just give everyone a wake up that in the final week you need to trust your most robust models and not necessarily those which rank the highest.
\nI have made it clear that the submission will not be useful in the private LB several times. So it will not affect your work or anyone else's in the final ranking, unless they choose to blindly fork and use submissions.
\nIf you really need to see your public ranking you can sort like above in your submission page.
", "rawMarkdown": "I have been on the other side of this in other competitions too, a public LB being wiped should just give everyone a wake up that in the final week you need to trust your most robust models and not necessarily those which rank the highest.\n\nI have made it clear that the submission will not be useful in the private LB several times. So it will not affect your work or anyone else's in the final ranking, unless they choose to blindly fork and use submissions.\n\nIf you really need to see your public ranking you can sort like above in your submission page.", "votes": -7 } ] }, { "id": 1912651, "postDate": "2022-08-24T22:24:05.387Z", "content": "I have 0.805 with blend ensemble of 6 LGBM models (0.794-0.796 scoring separately) on CV at home but only 0.797 on leaderboard….
\nWhy?
", "rawMarkdown": "I have 0.805 with blend ensemble of 6 LGBM models (0.794-0.796 scoring separately) on CV at home but only 0.797 on leaderboard….\n\nWhy?" }, { "id": 1903158, "postDate": "2022-08-17T06:45:56.660Z", "content": "Will it be an overfitted model???
", "rawMarkdown": "Will it be an overfitted model???" }, { "id": 1901214, "postDate": "2022-08-16T14:08:06.013Z", "content": "I didn't see this 0.8 code, was it deleted?
\nThis competition is really weird, I hope this 0.799 public score is really overfitted because I guess 90% of the medals now are people who simply resubmitted that score.
Don't worry too much about this, it will distort the predictions in the private dataset and their private predictions will get worse.
", "rawMarkdown": "Don't worry too much about this, it will distort the predictions in the private dataset and their private predictions will get worse." }, { "id": 1900161, "postDate": "2022-08-15T18:42:07.507Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 1900172, "postDate": "2022-08-15T18:49:43.447Z", "content": "And here is the main character of today..\"master of discussions\" won't give me anything..
\nThe only question is - why are you leading an army of copy-masters?
\nBreak the system? Laugh? Feel yourself in the center of the event?
I totally agree, this ruins the competitive spirit of many competitions for ~90% of participants. However, I am still unsure how to deal with this since especially beginners can learn a lot from public notebooks. Maybe the block for publishing top-scoring notebooks should come in earlier.
", "votes": 7, "replies": [] }, { "id": 1900164, "author_name": "Aaditya Agnihotri", "author_url": "", "post_date": "2022-08-15T18:46:10.853000", "content": "I definitely agree. Having read about what happened here, I am appalled at this behavior. While this may be a competition, it is meant for people to test their skills, not engage in what is borderline sabotage. One of my favorite part of Kaggle is how smart, yet incredibly helpful the community is, especially towards someone like me who doesn't have much experience.
", "votes": 1, "replies": [ { "id": 1900170, "author_name": "JM", "author_url": "", "post_date": "2022-08-15T18:49:35.667000", "content": "The question is, why would you use a public submission file if you are here to test your skills? Should you not be worrying about your own work? Trust your CV not the public LB?
\nI think you and OP seem to not be aware only the private LB is what matters..?
\nHere is a little trick to help you out if you deeply care about overfitting to the public LB:
\n
@julianmukaj You probably don't understand at all why everyone is so angry at you… competition as such is an excellent motivator for mastering new knowledge. And people like you discourage this motivation from less powerful participants.
", "votes": 6, "replies": [] }, { "id": 1900194, "author_name": "JM", "author_url": "", "post_date": "2022-08-15T19:08:42.170000", "content": "I have been on the other side of this in other competitions too, a public LB being wiped should just give everyone a wake up that in the final week you need to trust your most robust models and not necessarily those which rank the highest.
\nI have made it clear that the submission will not be useful in the private LB several times. So it will not affect your work or anyone else's in the final ranking, unless they choose to blindly fork and use submissions.
\nIf you really need to see your public ranking you can sort like above in your submission page.
", "votes": -7, "replies": [] } ] }, { "id": 1912651, "author_name": "Praise the Lord Jesus", "author_url": "", "post_date": "2022-08-24T22:24:05.387000", "content": "I have 0.805 with blend ensemble of 6 LGBM models (0.794-0.796 scoring separately) on CV at home but only 0.797 on leaderboard….
\nWhy?
", "votes": 0, "replies": [] }, { "id": 1903158, "author_name": "Nelson F. Rodriguez", "author_url": "", "post_date": "2022-08-17T06:45:56.660000", "content": "Will it be an overfitted model???
", "votes": 0, "replies": [] }, { "id": 1901214, "author_name": "Paulo Alberto Buoro", "author_url": "", "post_date": "2022-08-16T14:08:06.013000", "content": "I didn't see this 0.8 code, was it deleted?
\nThis competition is really weird, I hope this 0.799 public score is really overfitted because I guess 90% of the medals now are people who simply resubmitted that score.
Don't worry too much about this, it will distort the predictions in the private dataset and their private predictions will get worse.
", "votes": 0, "replies": [] }, { "id": 1900161, "author_name": "", "author_url": "", "post_date": "2022-08-15T18:42:07.507000", "content": "", "votes": 0, "replies": [ { "id": 1900172, "author_name": "Dmitry Uarov", "author_url": "", "post_date": "2022-08-15T18:49:43.447000", "content": "And here is the main character of today..\"master of discussions\" won't give me anything..
\nThe only question is - why are you leading an army of copy-masters?
\nBreak the system? Laugh? Feel yourself in the center of the event?