![\"\"](\"https://i.imgur.com/EyVUwX9.png\")
#10: Same country as #5. Only follows (mutually) #5.
\n#9: Same city as #5.
\n#8: Same city as #5. Ex-teammate.
It is likely that there is some private sharing going on. Apologies if it is a false alarm but it could be a very big coincidence.
", "messages": [ { "id": "1179116", "postDate": "01/31/2021 10:28:22", "content": "
#10: Same country as #5. Only follows (mutually) #5.
\n#9: Same city as #5.
\n#8: Same city as #5. Ex-teammate.
It is likely that there is some private sharing going on. Apologies if it is a false alarm but it could be a very big coincidence.
", "rawMarkdown": "\n\n**#10**: Same country as **#5**. Only follows (mutually) **#5**.\n**#9**: Same city as **#5**.\n**#8**: Same city as **#5**. Ex-teammate.\n\nIt is likely that there is some private sharing going on. Apologies if it is a false alarm but it could be a very big coincidence.", "votes": null }, { "id": "1179198", "postDate": "01/31/2021 11:57:49", "content": "I also saw the #8 & #5 connection a few days back. Also same company. To big of a coincidence for my opinion.
", "rawMarkdown": "I also saw the #8 & #5 connection a few days back. Also same company. To big of a coincidence for my opinion.", "votes": null }, { "id": "1179234", "postDate": "01/31/2021 12:42:36", "content": "I cannot tell for #9 for sure, but the other 3 work at the same firm.
\nLooks like obvious private sharing to me ^^
", "rawMarkdown": "I cannot tell for #9 for sure, but the other 3 work at the same firm. \n\nLooks like obvious private sharing to me ^^", "votes": null }, { "id": "1179291", "postDate": "01/31/2021 13:38:07", "content": "So what if they now merge? I agree that this looks very suspicious.
", "rawMarkdown": "So what if they now merge? I agree that this looks very suspicious.", "votes": null }, { "id": "1179389", "postDate": "01/31/2021 14:32:58", "content": "It looks suspicious.
", "rawMarkdown": "It looks suspicious.", "votes": null }, { "id": "1179393", "postDate": "01/31/2021 14:33:48", "content": "I noticed some of it as well. Good you created a topic.
", "rawMarkdown": "I noticed some of it as well. Good you created a topic.", "votes": null }, { "id": "1179451", "postDate": "01/31/2021 15:18:48", "content": "Maybe most birds are living in Texas near some quantum labs
", "rawMarkdown": "Maybe most birds are living in Texas near some quantum labs", "votes": null }, { "id": "1179854", "postDate": "01/31/2021 22:23:33", "content": "This reminds me of Saul Goodman's quote \" I know a Guy who Knows a Guy \".
", "rawMarkdown": "This reminds me of Saul Goodman's quote \" I know a Guy who Knows a Guy \".", "votes": null }, { "id": "1179908", "postDate": "02/01/2021 00:48:38", "content": "I work in same company as Ahmet who created the topic, and Dieter who is above me in LB, but it does not imply that we are sharing info. I can tell you they'd love to know how I could get 0.931 at my first sub :D . Similarly there are several H2O employees in top 50. I don't think they share when they are not in the same team.
\nBut in this case I also think there is private sharing. They all land in LB at same high level with very few subs. I hope for them, and for us, that they will all merge in a single team soon.
", "rawMarkdown": "I work in same company as Ahmet who created the topic, and Dieter who is above me in LB, but it does not imply that we are sharing info. I can tell you they'd love to know how I could get 0.931 at my first sub :D . Similarly there are several H2O employees in top 50. I don't think they share when they are not in the same team.\n\nBut in this case I also think there is private sharing. They all land in LB at same high level with very few subs. I hope for them, and for us, that they will all merge in a single team soon.", "votes": null }, { "id": "1179910", "postDate": "02/01/2021 00:50:57", "content": "It's great that there are detectives like you who guard the fairness of Kaggle competition.
\nI appreciate that, especially in times I am more behind in the leader board.
We are working together as a team. We will merge later before the final week.
\nWe work in this way so that we can make use of the unused submission quota allowed for a team.
Is there any other concern?
", "rawMarkdown": "It's great that there are detectives like you who guard the fairness of Kaggle competition.\nI appreciate that, especially in times I am more behind in the leader board.\n\nWe are working together as a team. We will merge later before the final week. \nWe work in this way so that we can make use of the unused submission quota allowed for a team.\n\nIs there any other concern?", "votes": null }, { "id": "1179912", "postDate": "02/01/2021 00:56:49", "content": "Similar topic I posted a month ago and got downvoted and mocked hard
\nhttps://www.kaggle.com/c/rfcx-species-audio-detection/discussion/208577
Thanks for making it clear.
\nI share your view that it is fine if you end up teaming, but some disagree. Indeed, rules say it is forbidden to share outside a team. You are not yet a team hence one can interpret rules as saying you should not share.
\nThis said I am not from Kaggle, not linked to Kaggle in anyway, and the above is just my opinion. Maybe you should contact Kaggle staff if you have any doubt.
\nGood luck anyway, this competition is funnier and funnier as it goes. It is the first time in a while that random modification of a public notebook isn't enough to get you in gold zone. It shows by the spread of LB scores. Let the best win!
\nEdit. A bit of context may help. We all suspect that a similar thing happened with CV experts form X5 company in Russia. Except they did not team in the end. In several comps they had same score up to the last shown digit. Since then people are nervous about this, hence this whole topic.
", "rawMarkdown": "Thanks for making it clear.\n\nI share your view that it is fine if you end up teaming, but some disagree. Indeed, rules say it is forbidden to share outside a team. You are not yet a team hence one can interpret rules as saying you should not share. \n\nThis said I am not from Kaggle, not linked to Kaggle in anyway, and the above is just my opinion. Maybe you should contact Kaggle staff if you have any doubt.\n\nGood luck anyway, this competition is funnier and funnier as it goes. It is the first time in a while that random modification of a public notebook isn't enough to get you in gold zone. It shows by the spread of LB scores. Let the best win!\n\nEdit. A bit of context may help. We all suspect that a similar thing happened with CV experts form X5 company in Russia. Except they did not team in the end. In several comps they had same score up to the last shown digit. Since then people are nervous about this, hence this whole topic.", "votes": null }, { "id": "1179916", "postDate": "02/01/2021 01:10:37", "content": "It's not clear to me why you think merging later makes sense. My reasoning is as follows:
\nSo any sharing without forming a team constitutes a violation of competition rules.
\nPerhaps you can share your interpretation of rules.
@cpmpml Thanks for bringing in the context. It makes more sense to me now why people has a much stronger reaction than I would expect. According to your EDIT, It seems someone who was previously private sharing in the finalized leader board was not penalized, which would definitely hurt the feelings of hard working Kagglers and breach the competition fairness. I would expect people who privately shares at the end to be found out easily like this and their scores to be invalidated. Someone who plans to private share at the end would do better camouflage than this.
\n@meaninglesslives My interpretation of the rule is that private sharing is bad since it hurts other Kagglers' final ranks and makes Kaggle medals less trustworthy. In this sense, what I did won't do any harm.
\nBut it's good that you guys point out that in the rule
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.
It is true that we are not technically a team yet. We will merge right now.
\nI apologize for the bad feelings and associations I might have caused.
", "rawMarkdown": "cpmpml Thanks for bringing in the context. It makes more sense to me now why people has a much stronger reaction than I would expect. According to your EDIT, It seems someone who was previously private sharing in the finalized leader board was not penalized, which would definitely hurt the feelings of hard working Kagglers and breach the competition fairness. I would expect people who privately shares at the end to be found out easily like this and their scores to be invalidated. Someone who plans to private share at the end would do better camouflage than this.\n\n@meaninglesslives My interpretation of the rule is that private sharing is bad since it hurts other Kagglers' final ranks and makes Kaggle medals less trustworthy. In this sense, what I did won't do any harm. \n\nBut it's good that you guys point out that in the rule\n> No private sharing outside teams\nPrivately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.\n\nIt is true that we are not technically a team yet. We will merge right now.\n\nI apologize for the bad feelings and associations I might have caused.", "votes": null }, { "id": "1179934", "postDate": "02/01/2021 02:04:15", "content": "I think it all comes to fairness. By not merging you can use more submissions per day than others. It could be seen as unfair. I personally don't care, as LB climbing is not how I see Kaggle competitions, but this is not a widely shared opinion ;)
", "rawMarkdown": "I think it all comes to fairness. By not merging you can use more submissions per day than others. It could be seen as unfair. I personally don't care, as LB climbing is not how I see Kaggle competitions, but this is not a widely shared opinion ;)", "votes": null }, { "id": "1180056", "postDate": "02/01/2021 04:21:41", "content": "I have thought it doesn't really matter whether people will team up in the future or not - if people take part in private sharing \"now\" it is against the rules.
", "rawMarkdown": "I have thought it doesn't really matter whether people will team up in the future or not - if people take part in private sharing \"now\" it is against the rules.", "votes": null }, { "id": "1180146", "postDate": "02/01/2021 05:57:58", "content": "We work in this way so that we can make use of the unused submission quota allowed for a team.
\nThats one reason why sharing outside a team is forbidden. I would not be surprised if you and your mates will be removed from LB after the competition. I am even in favour of your team being removed. In this competition where it is very hard to built a good cross-validation cheating to get more submissions has even more impact than in other competitions.
I teamed up with them (yesterday) before this thread was in limelight, I am not part of this \"incident\". I worked hard from the beginning of this competition slowly rising up the LB with 85+ subs. I had also contacted fellow kaggle-mates for teaming up ( @theoviel , @ybabakhin , @kneroma know about that) earlier when I was in silver zone.
\nNow since everyone has merge together the problem of more than average submission is resolved, for which most of the people were concerned about & its still 2 weeks for the competition to end.
\nCompetition host @jacklebien & @zephyrgold please provide a answer for everyone.
\n@juliaelliott @inversion
Regardless of what happens here, it is one further thing where I hope for Kaggle to make a clear statement about the rules.
", "rawMarkdown": "Regardless of what happens here, it is one further thing where I hope for Kaggle to make a clear statement about the rules.", "votes": null }, { "id": "1180436", "postDate": "02/01/2021 08:53:21", "content": "Although your claim was justified, their submission was public so anyone could've forked it and improve it a bit :)
", "rawMarkdown": "Although your claim was justified, their submission was public so anyone could've forked it and improve it a bit :)", "votes": null }, { "id": "1180452", "postDate": "02/01/2021 09:06:08", "content": "\n\nhttps://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/discussion/193632
\n
\n https://www.kaggle.com/c/abstraction-and-reasoning-challenge/discussion/150294
\n https://www.kaggle.com/c/landmark-retrieval-2020/discussion/170773
The last three \"exposed\" teams I know of. First two got removed, last one didn't (probably because they still merged early in the competition, I don't know).
\nThey must have some rule-based decision algorithm but I don't have enough data to reverse-engineer it ^^
", "rawMarkdown": "> https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/discussion/193632\n> https://www.kaggle.com/c/abstraction-and-reasoning-challenge/discussion/150294\n> https://www.kaggle.com/c/landmark-retrieval-2020/discussion/170773\n\nThe last three \"exposed\" teams I know of. First two got removed, last one didn't (probably because they still merged early in the competition, I don't know). \n\nThey must have some rule-based decision algorithm but I don't have enough data to reverse-engineer it ^^", "votes": null }, { "id": "1180461", "postDate": "02/01/2021 09:11:34", "content": "wow thanks for creating this topic i am so shocked that people are private sharing there data regardless of the competition rules . @juliaelliott , @zephyrgold , @jacklebien please take a good look at this situation .
", "rawMarkdown": "wow thanks for creating this topic i am so shocked that people are private sharing there data regardless of the competition rules . @juliaelliott , @zephyrgold , @jacklebien please take a good look at this situation .", "votes": null }, { "id": "1180463", "postDate": "02/01/2021 09:11:56", "content": "Merging later means that people who privately share have access to more submissions than the competitors who are competing fairly. There is a clear advantage to that.
", "rawMarkdown": "Merging later means that people who privately share have access to more submissions than the competitors who are competing fairly. There is a clear advantage to that.", "votes": null }, { "id": "1180622", "postDate": "02/01/2021 11:30:17", "content": "Not merging to have more submission slots is definitely shady.
\nProbably I would not vote for LB removal but that is not our decision…
I was just completing Ahmet's observations with what I found with a bit of stalking, not implying that people that work in the same companies share private infos ;)
", "rawMarkdown": "I was just completing Ahmet's observations with what I found with a bit of stalking, not implying that people that work in the same companies share private infos ;)", "votes": null }, { "id": "1180693", "postDate": "02/01/2021 12:20:34", "content": "Ahmet, your post forced a clarification and a team merge. This is awesome. The competition is fairer now thanks to you.
", "rawMarkdown": "Ahmet, your post forced a clarification and a team merge. This is awesome. The competition is fairer now thanks to you.", "votes": null }, { "id": "1180778", "postDate": "02/01/2021 13:14:51", "content": "I agree.
\nIf anything happen to me and other member of my team, I hope @rsinda would be free from any bad influence by this incidence. As he mentioned, we merged yesterday before which we didn't talk at all.
", "rawMarkdown": "I agree. \n\nIf anything happen to me and other member of my team, I hope @rsinda would be free from any bad influence by this incidence. As he mentioned, we merged yesterday before which we didn't talk at all.", "votes": null }, { "id": "1180849", "postDate": "02/01/2021 14:01:48", "content": "Thanks for stating your concerns here!
\nAs for the more submissions we have, I would say, we did get more flexibility in when to use the submissions, which is an advantage over others, in the sense that if 5 submissions per day for a team is not used up, we can use it later to some extent, however, we didn't have more total submission allowance than any other team. Per the competition rule
\nTeam Mergers\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\nIf our total submission is more than allowed by a team, then, we won't be able to merge. Same company, friend relationship. location, and past collaboration records would reveal our actual teamwork in the final leader board, if not easier than now, then, almost definitely, a well defined private sharing case will result in removal of all of our scores.
\nSo, we count the number of total submissions so that it does not exceed what is allowed for a single team. As you can see, after we have merged now, even with a new member that was not in the actually team before, the total number of submission is still well below the maximum submission allowed.
\nYes, I agree, if I think again, submission allowance later in the competition is more valuable than earlier in the competition, since we have more understanding of the competition, and the competitive nature of competition is more intense. What I did is to exchange early submission quota to later submission quota, which works in my favor.
\nIn short, we didn't have more submissions, but we did get an edge of submission flexibility by not merging early. I thought it is fine since we are actually a team. Now that I know better of the rule, the attitude of other participants, I wouldn't do that.
", "rawMarkdown": "Thanks for stating your concerns here!\n\nAs for the more submissions we have, I would say, we did get more flexibility in when to use the submissions, which is an advantage over others, in the sense that if 5 submissions per day for a team is not used up, we can use it later to some extent, however, we didn't have more total submission allowance than any other team. Per the competition rule \n```\nTeam Mergers\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\n```\n\nIf our total submission is more than allowed by a team, then, we won't be able to merge. Same company, friend relationship. location, and past collaboration records would reveal our actual teamwork in the final leader board, if not easier than now, then, almost definitely, a well defined private sharing case will result in removal of all of our scores.\n\nSo, we count the number of total submissions so that it does not exceed what is allowed for a single team. As you can see, after we have merged now, even with a new member that was not in the actually team before, the total number of submission is still well below the maximum submission allowed.\n\nYes, I agree, if I think again, submission allowance later in the competition is more valuable than earlier in the competition, since we have more understanding of the competition, and the competitive nature of competition is more intense. What I did is to exchange early submission quota to later submission quota, which works in my favor. \n\nIn short, we didn't have more submissions, but we did get an edge of submission flexibility by not merging early. I thought it is fine since we are actually a team. Now that I know better of the rule, the attitude of other participants, I wouldn't do that.", "votes": null }, { "id": "1182814", "postDate": "02/02/2021 15:55:47", "content": "Looks interesting!
", "rawMarkdown": "Looks interesting!", "votes": null }, { "id": "1183035", "postDate": "02/02/2021 17:48:23", "content": "@inversion @juliaelliott @zephyrgold
", "rawMarkdown": "inversion @juliaelliott @zephyrgold", "votes": null }, { "id": "1183357", "postDate": "02/02/2021 23:24:57", "content": "@barnwellguy ok but if you want to maximize the submission counts, you should join the competition earlier instead of merging later😿
", "rawMarkdown": "barnwellguy ok but if you want to maximize the submission counts, you should join the competition earlier instead of merging later😿", "votes": null } ], "comments": [ { "id": 1179198, "author_name": "christofhenkel", "author_url": "", "post_date": "01/31/2021 11:57:49", "content": "I also saw the #8 & #5 connection a few days back. Also same company. To big of a coincidence for my opinion.
", "votes": null, "replies": [] }, { "id": 1179234, "author_name": "theoviel", "author_url": "", "post_date": "01/31/2021 12:42:36", "content": "I cannot tell for #9 for sure, but the other 3 work at the same firm.
\nLooks like obvious private sharing to me ^^
", "votes": null, "replies": [ { "id": 1179908, "author_name": "cpmpml", "author_url": "", "post_date": "02/01/2021 00:48:38", "content": "I work in same company as Ahmet who created the topic, and Dieter who is above me in LB, but it does not imply that we are sharing info. I can tell you they'd love to know how I could get 0.931 at my first sub :D . Similarly there are several H2O employees in top 50. I don't think they share when they are not in the same team.
\nBut in this case I also think there is private sharing. They all land in LB at same high level with very few subs. I hope for them, and for us, that they will all merge in a single team soon.
", "votes": null, "replies": [] }, { "id": 1180686, "author_name": "theoviel", "author_url": "", "post_date": "02/01/2021 12:17:03", "content": "I was just completing Ahmet's observations with what I found with a bit of stalking, not implying that people that work in the same companies share private infos ;)
", "votes": null, "replies": [] } ] }, { "id": 1179291, "author_name": "philippsinger", "author_url": "", "post_date": "01/31/2021 13:38:07", "content": "So what if they now merge? I agree that this looks very suspicious.
", "votes": null, "replies": [] }, { "id": 1179389, "author_name": "atamazian", "author_url": "", "post_date": "01/31/2021 14:32:58", "content": "It looks suspicious.
", "votes": null, "replies": [] }, { "id": 1179393, "author_name": "cpmpml", "author_url": "", "post_date": "01/31/2021 14:33:48", "content": "I noticed some of it as well. Good you created a topic.
", "votes": null, "replies": [] }, { "id": 1179451, "author_name": "mcggood", "author_url": "", "post_date": "01/31/2021 15:18:48", "content": "Maybe most birds are living in Texas near some quantum labs
", "votes": null, "replies": [] }, { "id": 1179854, "author_name": "sayedathar11", "author_url": "", "post_date": "01/31/2021 22:23:33", "content": "This reminds me of Saul Goodman's quote \" I know a Guy who Knows a Guy \".
", "votes": null, "replies": [] }, { "id": 1179910, "author_name": "barnwellguy", "author_url": "", "post_date": "02/01/2021 00:50:57", "content": "It's great that there are detectives like you who guard the fairness of Kaggle competition.
\nI appreciate that, especially in times I am more behind in the leader board.
We are working together as a team. We will merge later before the final week.
\nWe work in this way so that we can make use of the unused submission quota allowed for a team.
Is there any other concern?
", "votes": null, "replies": [ { "id": 1179913, "author_name": "cpmpml", "author_url": "", "post_date": "02/01/2021 00:58:53", "content": "Thanks for making it clear.
\nI share your view that it is fine if you end up teaming, but some disagree. Indeed, rules say it is forbidden to share outside a team. You are not yet a team hence one can interpret rules as saying you should not share.
\nThis said I am not from Kaggle, not linked to Kaggle in anyway, and the above is just my opinion. Maybe you should contact Kaggle staff if you have any doubt.
\nGood luck anyway, this competition is funnier and funnier as it goes. It is the first time in a while that random modification of a public notebook isn't enough to get you in gold zone. It shows by the spread of LB scores. Let the best win!
\nEdit. A bit of context may help. We all suspect that a similar thing happened with CV experts form X5 company in Russia. Except they did not team in the end. In several comps they had same score up to the last shown digit. Since then people are nervous about this, hence this whole topic.
", "votes": null, "replies": [] }, { "id": 1179916, "author_name": "meaninglesslives", "author_url": "", "post_date": "02/01/2021 01:10:37", "content": "It's not clear to me why you think merging later makes sense. My reasoning is as follows:
\nSo any sharing without forming a team constitutes a violation of competition rules.
\nPerhaps you can share your interpretation of rules.
@cpmpml Thanks for bringing in the context. It makes more sense to me now why people has a much stronger reaction than I would expect. According to your EDIT, It seems someone who was previously private sharing in the finalized leader board was not penalized, which would definitely hurt the feelings of hard working Kagglers and breach the competition fairness. I would expect people who privately shares at the end to be found out easily like this and their scores to be invalidated. Someone who plans to private share at the end would do better camouflage than this.
\n@meaninglesslives My interpretation of the rule is that private sharing is bad since it hurts other Kagglers' final ranks and makes Kaggle medals less trustworthy. In this sense, what I did won't do any harm.
\nBut it's good that you guys point out that in the rule
\n\n\nNo private sharing outside teams
\n
\n Privately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.
It is true that we are not technically a team yet. We will merge right now.
\nI apologize for the bad feelings and associations I might have caused.
", "votes": null, "replies": [] }, { "id": 1179934, "author_name": "cpmpml", "author_url": "", "post_date": "02/01/2021 02:04:15", "content": "I think it all comes to fairness. By not merging you can use more submissions per day than others. It could be seen as unfair. I personally don't care, as LB climbing is not how I see Kaggle competitions, but this is not a widely shared opinion ;)
", "votes": null, "replies": [] }, { "id": 1180056, "author_name": "artgor", "author_url": "", "post_date": "02/01/2021 04:21:41", "content": "I have thought it doesn't really matter whether people will team up in the future or not - if people take part in private sharing \"now\" it is against the rules.
", "votes": null, "replies": [] }, { "id": 1180146, "author_name": "christofhenkel", "author_url": "", "post_date": "02/01/2021 05:57:58", "content": "We work in this way so that we can make use of the unused submission quota allowed for a team.
\nThats one reason why sharing outside a team is forbidden. I would not be surprised if you and your mates will be removed from LB after the competition. I am even in favour of your team being removed. In this competition where it is very hard to built a good cross-validation cheating to get more submissions has even more impact than in other competitions.
Regardless of what happens here, it is one further thing where I hope for Kaggle to make a clear statement about the rules.
", "votes": null, "replies": [] }, { "id": 1180452, "author_name": "theoviel", "author_url": "", "post_date": "02/01/2021 09:06:08", "content": "\n\nhttps://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/discussion/193632
\n
\n https://www.kaggle.com/c/abstraction-and-reasoning-challenge/discussion/150294
\n https://www.kaggle.com/c/landmark-retrieval-2020/discussion/170773
The last three \"exposed\" teams I know of. First two got removed, last one didn't (probably because they still merged early in the competition, I don't know).
\nThey must have some rule-based decision algorithm but I don't have enough data to reverse-engineer it ^^
", "votes": null, "replies": [] }, { "id": 1180463, "author_name": "anjum48", "author_url": "", "post_date": "02/01/2021 09:11:56", "content": "Merging later means that people who privately share have access to more submissions than the competitors who are competing fairly. There is a clear advantage to that.
", "votes": null, "replies": [] }, { "id": 1180622, "author_name": "gaborfodor", "author_url": "", "post_date": "02/01/2021 11:30:17", "content": "Not merging to have more submission slots is definitely shady.
\nProbably I would not vote for LB removal but that is not our decision…
Thanks for stating your concerns here!
\nAs for the more submissions we have, I would say, we did get more flexibility in when to use the submissions, which is an advantage over others, in the sense that if 5 submissions per day for a team is not used up, we can use it later to some extent, however, we didn't have more total submission allowance than any other team. Per the competition rule
\nTeam Mergers\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\nIf our total submission is more than allowed by a team, then, we won't be able to merge. Same company, friend relationship. location, and past collaboration records would reveal our actual teamwork in the final leader board, if not easier than now, then, almost definitely, a well defined private sharing case will result in removal of all of our scores.
\nSo, we count the number of total submissions so that it does not exceed what is allowed for a single team. As you can see, after we have merged now, even with a new member that was not in the actually team before, the total number of submission is still well below the maximum submission allowed.
\nYes, I agree, if I think again, submission allowance later in the competition is more valuable than earlier in the competition, since we have more understanding of the competition, and the competitive nature of competition is more intense. What I did is to exchange early submission quota to later submission quota, which works in my favor.
\nIn short, we didn't have more submissions, but we did get an edge of submission flexibility by not merging early. I thought it is fine since we are actually a team. Now that I know better of the rule, the attitude of other participants, I wouldn't do that.
", "votes": null, "replies": [] }, { "id": 1183357, "author_name": "underwearfitting", "author_url": "", "post_date": "02/02/2021 23:24:57", "content": "@barnwellguy ok but if you want to maximize the submission counts, you should join the competition earlier instead of merging later😿
", "votes": null, "replies": [] } ] }, { "id": 1179912, "author_name": "jihangz", "author_url": "", "post_date": "02/01/2021 00:56:49", "content": "Similar topic I posted a month ago and got downvoted and mocked hard
\nhttps://www.kaggle.com/c/rfcx-species-audio-detection/discussion/208577
Although your claim was justified, their submission was public so anyone could've forked it and improve it a bit :)
", "votes": null, "replies": [] } ] }, { "id": 1180156, "author_name": "rsinda", "author_url": "", "post_date": "02/01/2021 06:08:10", "content": "I teamed up with them (yesterday) before this thread was in limelight, I am not part of this \"incident\". I worked hard from the beginning of this competition slowly rising up the LB with 85+ subs. I had also contacted fellow kaggle-mates for teaming up ( @theoviel , @ybabakhin , @kneroma know about that) earlier when I was in silver zone.
\nNow since everyone has merge together the problem of more than average submission is resolved, for which most of the people were concerned about & its still 2 weeks for the competition to end.
\nCompetition host @jacklebien & @zephyrgold please provide a answer for everyone.
\n@juliaelliott @inversion
I agree.
\nIf anything happen to me and other member of my team, I hope @rsinda would be free from any bad influence by this incidence. As he mentioned, we merged yesterday before which we didn't talk at all.
", "votes": null, "replies": [] } ] }, { "id": 1180461, "author_name": "trooperog", "author_url": "", "post_date": "02/01/2021 09:11:34", "content": "wow thanks for creating this topic i am so shocked that people are private sharing there data regardless of the competition rules . @juliaelliott , @zephyrgold , @jacklebien please take a good look at this situation .
", "votes": null, "replies": [] }, { "id": 1180693, "author_name": "cpmpml", "author_url": "", "post_date": "02/01/2021 12:20:34", "content": "Ahmet, your post forced a clarification and a team merge. This is awesome. The competition is fairer now thanks to you.
", "votes": null, "replies": [] }, { "id": 1182814, "author_name": "kickitlikeshika", "author_url": "", "post_date": "02/02/2021 15:55:47", "content": "Looks interesting!
", "votes": null, "replies": [] }, { "id": 1183035, "author_name": "ks2019", "author_url": "", "post_date": "02/02/2021 17:48:23", "content": "", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1179116": "\n\n**#10**: Same country as **#5**. Only follows (mutually) **#5**.\n**#9**: Same city as **#5**.\n**#8**: Same city as **#5**. Ex-teammate.\n\nIt is likely that there is some private sharing going on. Apologies if it is a false alarm but it could be a very big coincidence.", "1179198": "I also saw the #8 & #5 connection a few days back. Also same company. To big of a coincidence for my opinion.", "1179234": "I cannot tell for #9 for sure, but the other 3 work at the same firm. \n\nLooks like obvious private sharing to me ^^", "1179291": "So what if they now merge? I agree that this looks very suspicious.", "1179389": "It looks suspicious.", "1179393": "I noticed some of it as well. Good you created a topic.", "1179451": "Maybe most birds are living in Texas near some quantum labs", "1179854": "This reminds me of Saul Goodman's quote \" I know a Guy who Knows a Guy \".", "1179908": "I work in same company as Ahmet who created the topic, and Dieter who is above me in LB, but it does not imply that we are sharing info. I can tell you they'd love to know how I could get 0.931 at my first sub :D . Similarly there are several H2O employees in top 50. I don't think they share when they are not in the same team.\n\nBut in this case I also think there is private sharing. They all land in LB at same high level with very few subs. I hope for them, and for us, that they will all merge in a single team soon.", "1179910": "It's great that there are detectives like you who guard the fairness of Kaggle competition.\nI appreciate that, especially in times I am more behind in the leader board.\n\nWe are working together as a team. We will merge later before the final week. \nWe work in this way so that we can make use of the unused submission quota allowed for a team.\n\nIs there any other concern?", "1179912": "Similar topic I posted a month ago and got downvoted and mocked hard\nhttps://www.kaggle.com/c/rfcx-species-audio-detection/discussion/208577", "1179913": "Thanks for making it clear.\n\nI share your view that it is fine if you end up teaming, but some disagree. Indeed, rules say it is forbidden to share outside a team. You are not yet a team hence one can interpret rules as saying you should not share. \n\nThis said I am not from Kaggle, not linked to Kaggle in anyway, and the above is just my opinion. Maybe you should contact Kaggle staff if you have any doubt.\n\nGood luck anyway, this competition is funnier and funnier as it goes. It is the first time in a while that random modification of a public notebook isn't enough to get you in gold zone. It shows by the spread of LB scores. Let the best win!\n\nEdit. A bit of context may help. We all suspect that a similar thing happened with CV experts form X5 company in Russia. Except they did not team in the end. In several comps they had same score up to the last shown digit. Since then people are nervous about this, hence this whole topic.", "1179916": "It's not clear to me why you think merging **later** makes sense. My reasoning is as follows:\n1. No private sharing outside teams (Competition Rules)\n2. At the present moment you are not a team.\n\nSo any sharing without forming a team constitutes a violation of competition rules.\nPerhaps you can share your interpretation of rules.", "1179928": "cpmpml Thanks for bringing in the context. It makes more sense to me now why people has a much stronger reaction than I would expect. According to your EDIT, It seems someone who was previously private sharing in the finalized leader board was not penalized, which would definitely hurt the feelings of hard working Kagglers and breach the competition fairness. I would expect people who privately shares at the end to be found out easily like this and their scores to be invalidated. Someone who plans to private share at the end would do better camouflage than this.\n\n@meaninglesslives My interpretation of the rule is that private sharing is bad since it hurts other Kagglers' final ranks and makes Kaggle medals less trustworthy. In this sense, what I did won't do any harm. \n\nBut it's good that you guys point out that in the rule\n> No private sharing outside teams\nPrivately sharing code or data outside of teams is not permitted. It's okay to share code if made available to all participants on the forums.\n\nIt is true that we are not technically a team yet. We will merge right now.\n\nI apologize for the bad feelings and associations I might have caused.", "1179934": "I think it all comes to fairness. By not merging you can use more submissions per day than others. It could be seen as unfair. I personally don't care, as LB climbing is not how I see Kaggle competitions, but this is not a widely shared opinion ;)", "1180056": "I have thought it doesn't really matter whether people will team up in the future or not - if people take part in private sharing \"now\" it is against the rules.", "1180146": "**We work in this way so that we can make use of the unused submission quota allowed for a team.**\nThats one reason why sharing outside a team is forbidden. I would not be surprised if you and your mates will be removed from LB after the competition. I am even in favour of your team being removed. In this competition where it is very hard to built a good cross-validation cheating to get more submissions has even more impact than in other competitions.", "1180156": "I teamed up with them (yesterday) before this thread was in limelight, I am not part of this \"incident\". I worked hard from the beginning of this competition slowly rising up the LB with 85+ subs. I had also contacted fellow kaggle-mates for teaming up ( @theoviel , @ybabakhin , @kneroma know about that) earlier when I was in silver zone. \n\nNow since everyone has merge together the problem of more than average submission is resolved, for which most of the people were concerned about & its still 2 weeks for the competition to end.\n\nCompetition host @jacklebien & @zephyrgold please provide a answer for everyone.\n@juliaelliott @inversion", "1180272": "Regardless of what happens here, it is one further thing where I hope for Kaggle to make a clear statement about the rules.", "1180436": "Although your claim was justified, their submission was public so anyone could've forked it and improve it a bit :)", "1180452": "> https://www.kaggle.com/c/rsna-str-pulmonary-embolism-detection/discussion/193632\n> https://www.kaggle.com/c/abstraction-and-reasoning-challenge/discussion/150294\n> https://www.kaggle.com/c/landmark-retrieval-2020/discussion/170773\n\nThe last three \"exposed\" teams I know of. First two got removed, last one didn't (probably because they still merged early in the competition, I don't know). \n\nThey must have some rule-based decision algorithm but I don't have enough data to reverse-engineer it ^^", "1180461": "wow thanks for creating this topic i am so shocked that people are private sharing there data regardless of the competition rules . @juliaelliott , @zephyrgold , @jacklebien please take a good look at this situation .", "1180463": "Merging later means that people who privately share have access to more submissions than the competitors who are competing fairly. There is a clear advantage to that.", "1180622": "Not merging to have more submission slots is definitely shady. \nProbably I would not vote for LB removal but that is not our decision...", "1180686": "I was just completing Ahmet's observations with what I found with a bit of stalking, not implying that people that work in the same companies share private infos ;)", "1180693": "Ahmet, your post forced a clarification and a team merge. This is awesome. The competition is fairer now thanks to you.", "1180778": "I agree. \n\nIf anything happen to me and other member of my team, I hope @rsinda would be free from any bad influence by this incidence. As he mentioned, we merged yesterday before which we didn't talk at all.", "1180849": "Thanks for stating your concerns here!\n\nAs for the more submissions we have, I would say, we did get more flexibility in when to use the submissions, which is an advantage over others, in the sense that if 5 submissions per day for a team is not used up, we can use it later to some extent, however, we didn't have more total submission allowance than any other team. Per the competition rule \n```\nTeam Mergers\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\n```\n\nIf our total submission is more than allowed by a team, then, we won't be able to merge. Same company, friend relationship. location, and past collaboration records would reveal our actual teamwork in the final leader board, if not easier than now, then, almost definitely, a well defined private sharing case will result in removal of all of our scores.\n\nSo, we count the number of total submissions so that it does not exceed what is allowed for a single team. As you can see, after we have merged now, even with a new member that was not in the actually team before, the total number of submission is still well below the maximum submission allowed.\n\nYes, I agree, if I think again, submission allowance later in the competition is more valuable than earlier in the competition, since we have more understanding of the competition, and the competitive nature of competition is more intense. What I did is to exchange early submission quota to later submission quota, which works in my favor. \n\nIn short, we didn't have more submissions, but we did get an edge of submission flexibility by not merging early. I thought it is fine since we are actually a team. Now that I know better of the rule, the attitude of other participants, I wouldn't do that.", "1182814": "Looks interesting!", "1183035": "inversion @juliaelliott @zephyrgold", "1183357": "barnwellguy ok but if you want to maximize the submission counts, you should join the competition earlier instead of merging later😿" }, "source": "meta" }