Kaggle admins and competition hosts, can you please address this issue? It's disheartening to see. There appear to be at least 2 fake accounts in this competition that are being used to blatantly cheat.
\nThe evidence is pretty clear.
\nThese are temporary fake accounts created by users to probe the LB.
\nThank you in advance!
", "messages": [ { "id": "1497874", "postDate": "08/31/2021 13:23:15", "content": "\nKaggle admins and competition hosts, can you please address this issue? It's disheartening to see. There appear to be at least 2 fake accounts in this competition that are being used to blatantly cheat.
\nThe evidence is pretty clear.
\nThese are temporary fake accounts created by users to probe the LB.
\nThank you in advance!
", "rawMarkdown": "juliaelliott \n\n**Kaggle admins and competition hosts, can you please address this issue? It's disheartening to see.** There appear to be at least 2 fake accounts in this competition that are being used to blatantly cheat.\n\n* [**pegggggggggg**](https://www.kaggle.com/pegggggggggg)\n * Joined 1 day ago\n * Has a score of **1.0** on the LB\n* [**yifffff12**](https://www.kaggle.com/yifffff12)\n * Joined 3 days ago\n * Has a score of **0.879** on the LB\n\n---\n\nThe evidence is pretty clear. \n\nThese are temporary fake accounts created by users to probe the LB. \n* Additional evidence is the repeating letters in the name, the buckets of 5 submissions at a time, the high public LB scores, and the fact that the two users always submit at the same time as each other.\n\n---\n\nThank you in advance!", "votes": null }, { "id": "1497896", "postDate": "08/31/2021 13:46:26", "content": "@dschettler8845 But how can one cheat a competition just by creating 2 fake accounts ?
", "rawMarkdown": "dschettler8845 But how can one cheat a competition just by creating 2 fake accounts ?", "votes": null }, { "id": "1497923", "postDate": "08/31/2021 14:09:37", "content": "Creating multiple accounts allows you to:
\nIt is also LITERALLY, the first rule in this competition:
\n\n", "rawMarkdown": "Creating multiple accounts allows you to:\n\n1. Submit more than the 5 a day limit *(generally to probe the LB... as these individual(s) have)*\n2. Use way more GPU/TPU than you would normally be allowed\n3. Use other resources more (datasets, etc.)\n4. Hide your own real LB score (this has been beneficial in other competitions)\n\n---\n\nIt is also **LITERALLY**, the first rule in this competition:\n\n> **\"One account per participant\"**\n\n> **\"You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\"**", "votes": null }, { "id": "1497937", "postDate": "08/31/2021 14:22:18", "content": "\"One account per participant\"
\n\"You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\"
\n
@dschettler8845 Ohh .. 🙁
", "rawMarkdown": "dschettler8845 Ohh .. 🙁", "votes": null }, { "id": "1498192", "postDate": "08/31/2021 18:23:30", "content": "there are more I am sure
", "rawMarkdown": "there are more I am sure", "votes": null }, { "id": "1498638", "postDate": "09/01/2021 05:30:51", "content": "Literally lb-probing without fake accounts is against the rules too ;)
", "rawMarkdown": "Literally lb-probing without fake accounts is against the rules too ;)", "votes": null }, { "id": "1499608", "postDate": "09/01/2021 18:51:53", "content": "Is lb probing really against the rules?
", "rawMarkdown": "Is lb probing really against the rules?", "votes": null }, { "id": "1499753", "postDate": "09/01/2021 22:24:12", "content": "\n\nAs a participant in the Competition, you agree 1) not to make copies of, or in any way redistribute, any of the data made available to you, 2) not to attempt to re-identify any personal information based on the data, 3) not to attempt to probe the test set's labels, and 4) to notify the Competition organizers of any personally identifiable information you encounter in the data by posting a message to the Kaggle community discussion forums.
\n
You can find this on competition rules page. Chapter 7 - Competition data.
", "rawMarkdown": "> **As a participant in the Competition, you agree** 1) not to make copies of, or in any way redistribute, any of the data made available to you, 2) not to attempt to re-identify any personal information based on the data, 3) **not to attempt to probe the test set's labels**, and 4) to notify the Competition organizers of any personally identifiable information you encounter in the data by posting a message to the Kaggle community discussion forums.\n\nYou can find this on competition rules page. Chapter 7 - Competition data.", "votes": null }, { "id": "1499758", "postDate": "09/01/2021 22:35:16", "content": "That’s w.r.t probing the hidden test set as far as I understood it. Anytime you base a submission choice on how well it did on the public LB you’re technically guilty of “probing” the public LB in my opinion.
\nPublic LB probing, to some extent, has always been a part of Kaggle competitions (beyond what I previously described). But probing the Private LB for any real information (other than volume) has generally been restricted.
\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.
\nI created a separate thread previously to discuss the probing topic. If you’d like to add in an opinion or commentary I’m sure it would be appreciated in that thread as well.
\nHope this clears things up.
", "rawMarkdown": "That’s w.r.t probing the hidden test set as far as I understood it. Anytime you base a submission choice on how well it did on the public LB you’re technically guilty of “probing” the public LB in my opinion.\n\nPublic LB probing, to some extent, has always been a part of Kaggle competitions (beyond what I previously described). But probing the Private LB for any real information (other than volume) has generally been restricted.\n\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.\n\nI created a separate thread previously to discuss the probing topic. If you’d like to add in an opinion or commentary I’m sure it would be appreciated in that thread as well.\n\nHope this clears things up.", "votes": null }, { "id": "1499811", "postDate": "09/02/2021 00:12:51", "content": "\n\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.
\nThese are temporary fake accounts created by users to probe the LB.
\n
Really? ;)
\n\n\nThat’s w.r.t probing the hidden test set as far as I understood it
\n
How can you imagine to do it? It's hidden and we have no lb feedback about it.
", "rawMarkdown": "> That being said, this thread is about cheating using multiple accounts… not the rules around probing in general.\n\n> These are temporary fake accounts created by users to probe the LB.\n\nReally? ;)\n\n> That’s w.r.t probing the hidden test set as far as I understood it\n\nHow can you imagine to do it? It's hidden and we have no lb feedback about it.", "votes": null }, { "id": "1499812", "postDate": "09/02/2021 00:14:51", "content": "\n\ntest/ - the test files, which use the same structure as train/; your task is to predict the MGMT_value for each subject in the test data. NOTE: the total size of the rerun test set (Public and Private) is ~5x the size of the Public test set
\n3) not to attempt to probe the test set's labels
\n
I think it's more clear now!
", "rawMarkdown": "> test/ - the test files, which use the same structure as train/; your task is to predict the MGMT_value for each subject in the test data. NOTE: the total size of the rerun **test set (Public and Private)** is ~5x the size of the Public test set\n\n> 3) not to attempt to probe **the test set's labels**\n\nI think it's more clear now!", "votes": null }, { "id": "1499853", "postDate": "09/02/2021 02:01:35", "content": "I appreciate the effort to clear this up. Thanks, @sggpls.
", "rawMarkdown": "I appreciate the effort to clear this up. Thanks, @sggpls.", "votes": null }, { "id": "1500845", "postDate": "09/02/2021 17:06:35", "content": "This is a violation of not just one or two, but so many rules, proper action should be taken against them after \"probe\" by kaggle team
", "rawMarkdown": "This is a violation of not just one or two, but so many rules, proper action should be taken against them after \"probe\" by kaggle team", "votes": null }, { "id": "1501290", "postDate": "09/03/2021 05:02:42", "content": "kaggle monitors ip and mac addresses to monitor activity, even browser, so they should be automatically eliminated no?
", "rawMarkdown": "kaggle monitors ip and mac addresses to monitor activity, even browser, so they should be automatically eliminated no?", "votes": null }, { "id": "1501364", "postDate": "09/03/2021 06:28:08", "content": "You mean @pegggggggggg is a fake account of @onodera?
\nI don't know, but he is the only account with LB 1.
Hi, actually he did not say it, and we don't assume it.
", "rawMarkdown": "Hi, actually he did not say it, and we don't assume it.", "votes": null }, { "id": "1502385", "postDate": "09/04/2021 07:41:18", "content": "@dschettler8845 For people who have a hand labelled dataset , which they can easily use for their validation as compared to people who do not have a hand labelled dataset will be stuck with only option to split the training dataset for train/val , this essentially gives you a edge until and unless someone publishes their hand labelled dataset .
\nAnd if people have a hand labelled test output can kaggle in anyway know if you used it in your cross validation ??
", "rawMarkdown": "dschettler8845 For people who have a hand labelled dataset , which they can easily use for their validation as compared to people who do not have a hand labelled dataset will be stuck with only option to split the training dataset for train/val , this essentially gives you a edge until and unless someone publishes their hand labelled dataset .\n\nAnd if people have a hand labelled test output can kaggle in anyway know if you used it in your cross validation ??", "votes": null }, { "id": "1502744", "postDate": "09/04/2021 15:52:01", "content": "One more issue to be considered, I sew few competition and dataset experts in the only beginner 30 Days of ML leaderboard. It is disheartening that those experts take away the deservedness of many who are just starting into ML.
", "rawMarkdown": "One more issue to be considered, I sew few competition and dataset experts in the only beginner 30 Days of ML leaderboard. It is disheartening that those experts take away the deservedness of many who are just starting into ML.", "votes": null }, { "id": "1502934", "postDate": "09/04/2021 19:03:08", "content": "yeah I know some punks who cheat
", "rawMarkdown": "yeah I know some punks who cheat", "votes": null }, { "id": "1503367", "postDate": "09/05/2021 09:49:10", "content": "informitive 🙏🙏
", "rawMarkdown": "informitive 🙏🙏", "votes": null }, { "id": "1506568", "postDate": "09/08/2021 11:00:55", "content": "**How they can get 1.0 on the LB? **
\nHow we can predict true MGMT for 87 test sample ?
Due to the way the contestants have been probing the public leaderboard, can we suggest to omit these batch for evaluation and solely rely on the private test batch for scoring. There will still be bias/advantage as they could have used those labels for training. Can we release the label of public dataset to be fair to the competition ?
", "rawMarkdown": "Due to the way the contestants have been probing the public leaderboard, can we suggest to omit these batch for evaluation and solely rely on the private test batch for scoring. There will still be bias/advantage as they could have used those labels for training. Can we release the label of public dataset to be fair to the competition ?", "votes": null }, { "id": "1546323", "postDate": "10/16/2021 03:05:48", "content": "100% agree
", "rawMarkdown": "100% agree", "votes": null } ], "comments": [ { "id": 1497896, "author_name": "akpmpr", "author_url": "", "post_date": "08/31/2021 13:46:26", "content": "@dschettler8845 But how can one cheat a competition just by creating 2 fake accounts ?
", "votes": null, "replies": [ { "id": 1497923, "author_name": "dschettler8845", "author_url": "", "post_date": "08/31/2021 14:09:37", "content": "Creating multiple accounts allows you to:
\nIt is also LITERALLY, the first rule in this competition:
\n\n", "votes": null, "replies": [] }, { "id": 1497937, "author_name": "akpmpr", "author_url": "", "post_date": "08/31/2021 14:22:18", "content": "\"One account per participant\"
\n\"You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\"
\n
@dschettler8845 Ohh .. 🙁
", "votes": null, "replies": [] }, { "id": 1503367, "author_name": "mohsennaderali", "author_url": "", "post_date": "09/05/2021 09:49:10", "content": "informitive 🙏🙏
", "votes": null, "replies": [] } ] }, { "id": 1498192, "author_name": "mrinath", "author_url": "", "post_date": "08/31/2021 18:23:30", "content": "there are more I am sure
", "votes": null, "replies": [] }, { "id": 1498638, "author_name": "sggpls", "author_url": "", "post_date": "09/01/2021 05:30:51", "content": "Literally lb-probing without fake accounts is against the rules too ;)
", "votes": null, "replies": [ { "id": 1499608, "author_name": "gwanghan", "author_url": "", "post_date": "09/01/2021 18:51:53", "content": "Is lb probing really against the rules?
", "votes": null, "replies": [] }, { "id": 1499753, "author_name": "sggpls", "author_url": "", "post_date": "09/01/2021 22:24:12", "content": "\n\nAs a participant in the Competition, you agree 1) not to make copies of, or in any way redistribute, any of the data made available to you, 2) not to attempt to re-identify any personal information based on the data, 3) not to attempt to probe the test set's labels, and 4) to notify the Competition organizers of any personally identifiable information you encounter in the data by posting a message to the Kaggle community discussion forums.
\n
You can find this on competition rules page. Chapter 7 - Competition data.
", "votes": null, "replies": [] }, { "id": 1499758, "author_name": "dschettler8845", "author_url": "", "post_date": "09/01/2021 22:35:16", "content": "That’s w.r.t probing the hidden test set as far as I understood it. Anytime you base a submission choice on how well it did on the public LB you’re technically guilty of “probing” the public LB in my opinion.
\nPublic LB probing, to some extent, has always been a part of Kaggle competitions (beyond what I previously described). But probing the Private LB for any real information (other than volume) has generally been restricted.
\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.
\nI created a separate thread previously to discuss the probing topic. If you’d like to add in an opinion or commentary I’m sure it would be appreciated in that thread as well.
\nHope this clears things up.
", "votes": null, "replies": [] }, { "id": 1499811, "author_name": "sggpls", "author_url": "", "post_date": "09/02/2021 00:12:51", "content": "\n\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.
\nThese are temporary fake accounts created by users to probe the LB.
\n
Really? ;)
\n\n\nThat’s w.r.t probing the hidden test set as far as I understood it
\n
How can you imagine to do it? It's hidden and we have no lb feedback about it.
", "votes": null, "replies": [] }, { "id": 1499812, "author_name": "sggpls", "author_url": "", "post_date": "09/02/2021 00:14:51", "content": "\n\ntest/ - the test files, which use the same structure as train/; your task is to predict the MGMT_value for each subject in the test data. NOTE: the total size of the rerun test set (Public and Private) is ~5x the size of the Public test set
\n3) not to attempt to probe the test set's labels
\n
I think it's more clear now!
", "votes": null, "replies": [] }, { "id": 1499853, "author_name": "dschettler8845", "author_url": "", "post_date": "09/02/2021 02:01:35", "content": "I appreciate the effort to clear this up. Thanks, @sggpls.
", "votes": null, "replies": [] } ] }, { "id": 1500845, "author_name": "keagle", "author_url": "", "post_date": "09/02/2021 17:06:35", "content": "This is a violation of not just one or two, but so many rules, proper action should be taken against them after \"probe\" by kaggle team
", "votes": null, "replies": [] }, { "id": 1501290, "author_name": "shummi39", "author_url": "", "post_date": "09/03/2021 05:02:42", "content": "kaggle monitors ip and mac addresses to monitor activity, even browser, so they should be automatically eliminated no?
", "votes": null, "replies": [] }, { "id": 1501364, "author_name": "xiaowangiiiii", "author_url": "", "post_date": "09/03/2021 06:28:08", "content": "You mean @pegggggggggg is a fake account of @onodera?
\nI don't know, but he is the only account with LB 1.
Hi, actually he did not say it, and we don't assume it.
", "votes": null, "replies": [] } ] }, { "id": 1502385, "author_name": "avikrams", "author_url": "", "post_date": "09/04/2021 07:41:18", "content": "@dschettler8845 For people who have a hand labelled dataset , which they can easily use for their validation as compared to people who do not have a hand labelled dataset will be stuck with only option to split the training dataset for train/val , this essentially gives you a edge until and unless someone publishes their hand labelled dataset .
\nAnd if people have a hand labelled test output can kaggle in anyway know if you used it in your cross validation ??
", "votes": null, "replies": [] }, { "id": 1502744, "author_name": "hrithikpurwar", "author_url": "", "post_date": "09/04/2021 15:52:01", "content": "One more issue to be considered, I sew few competition and dataset experts in the only beginner 30 Days of ML leaderboard. It is disheartening that those experts take away the deservedness of many who are just starting into ML.
", "votes": null, "replies": [] }, { "id": 1502934, "author_name": "trafless", "author_url": "", "post_date": "09/04/2021 19:03:08", "content": "yeah I know some punks who cheat
", "votes": null, "replies": [] }, { "id": 1506568, "author_name": "mohammadhosein1998", "author_url": "", "post_date": "09/08/2021 11:00:55", "content": "**How they can get 1.0 on the LB? **
\nHow we can predict true MGMT for 87 test sample ?
Due to the way the contestants have been probing the public leaderboard, can we suggest to omit these batch for evaluation and solely rely on the private test batch for scoring. There will still be bias/advantage as they could have used those labels for training. Can we release the label of public dataset to be fair to the competition ?
", "votes": null, "replies": [ { "id": 1546323, "author_name": "dschettler8845", "author_url": "", "post_date": "10/16/2021 03:05:48", "content": "100% agree
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1497874": "juliaelliott \n\n**Kaggle admins and competition hosts, can you please address this issue? It's disheartening to see.** There appear to be at least 2 fake accounts in this competition that are being used to blatantly cheat.\n\n* [**pegggggggggg**](https://www.kaggle.com/pegggggggggg)\n * Joined 1 day ago\n * Has a score of **1.0** on the LB\n* [**yifffff12**](https://www.kaggle.com/yifffff12)\n * Joined 3 days ago\n * Has a score of **0.879** on the LB\n\n---\n\nThe evidence is pretty clear. \n\nThese are temporary fake accounts created by users to probe the LB. \n* Additional evidence is the repeating letters in the name, the buckets of 5 submissions at a time, the high public LB scores, and the fact that the two users always submit at the same time as each other.\n\n---\n\nThank you in advance!", "1497896": "dschettler8845 But how can one cheat a competition just by creating 2 fake accounts ?", "1497923": "Creating multiple accounts allows you to:\n\n1. Submit more than the 5 a day limit *(generally to probe the LB... as these individual(s) have)*\n2. Use way more GPU/TPU than you would normally be allowed\n3. Use other resources more (datasets, etc.)\n4. Hide your own real LB score (this has been beneficial in other competitions)\n\n---\n\nIt is also **LITERALLY**, the first rule in this competition:\n\n> **\"One account per participant\"**\n\n> **\"You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\"**", "1497937": "dschettler8845 Ohh .. 🙁", "1498192": "there are more I am sure", "1498638": "Literally lb-probing without fake accounts is against the rules too ;)", "1499608": "Is lb probing really against the rules?", "1499753": "> **As a participant in the Competition, you agree** 1) not to make copies of, or in any way redistribute, any of the data made available to you, 2) not to attempt to re-identify any personal information based on the data, 3) **not to attempt to probe the test set's labels**, and 4) to notify the Competition organizers of any personally identifiable information you encounter in the data by posting a message to the Kaggle community discussion forums.\n\nYou can find this on competition rules page. Chapter 7 - Competition data.", "1499758": "That’s w.r.t probing the hidden test set as far as I understood it. Anytime you base a submission choice on how well it did on the public LB you’re technically guilty of “probing” the public LB in my opinion.\n\nPublic LB probing, to some extent, has always been a part of Kaggle competitions (beyond what I previously described). But probing the Private LB for any real information (other than volume) has generally been restricted.\n\nThat being said, this thread is about cheating using multiple accounts… not the rules around probing in general.\n\nI created a separate thread previously to discuss the probing topic. If you’d like to add in an opinion or commentary I’m sure it would be appreciated in that thread as well.\n\nHope this clears things up.", "1499811": "> That being said, this thread is about cheating using multiple accounts… not the rules around probing in general.\n\n> These are temporary fake accounts created by users to probe the LB.\n\nReally? ;)\n\n> That’s w.r.t probing the hidden test set as far as I understood it\n\nHow can you imagine to do it? It's hidden and we have no lb feedback about it.", "1499812": "> test/ - the test files, which use the same structure as train/; your task is to predict the MGMT_value for each subject in the test data. NOTE: the total size of the rerun **test set (Public and Private)** is ~5x the size of the Public test set\n\n> 3) not to attempt to probe **the test set's labels**\n\nI think it's more clear now!", "1499853": "I appreciate the effort to clear this up. Thanks, @sggpls.", "1500845": "This is a violation of not just one or two, but so many rules, proper action should be taken against them after \"probe\" by kaggle team", "1501290": "kaggle monitors ip and mac addresses to monitor activity, even browser, so they should be automatically eliminated no?", "1501364": "You mean @pegggggggggg is a fake account of @onodera?\nI don't know, but he is the only account with LB 1.", "1502311": "Hi, actually he did not say it, and we don't assume it.", "1502385": "dschettler8845 For people who have a hand labelled dataset , which they can easily use for their validation as compared to people who do not have a hand labelled dataset will be stuck with only option to split the training dataset for train/val , this essentially gives you a edge until and unless someone publishes their hand labelled dataset .\n\nAnd if people have a hand labelled test output can kaggle in anyway know if you used it in your cross validation ??", "1502744": "One more issue to be considered, I sew few competition and dataset experts in the only beginner 30 Days of ML leaderboard. It is disheartening that those experts take away the deservedness of many who are just starting into ML.", "1502934": "yeah I know some punks who cheat", "1503367": "informitive 🙏🙏", "1506568": "**How they can get 1.0 on the LB? **\nHow we can predict true MGMT for 87 test sample ?", "1546305": "Due to the way the contestants have been probing the public leaderboard, can we suggest to omit these batch for evaluation and solely rely on the private test batch for scoring. There will still be bias/advantage as they could have used those labels for training. Can we release the label of public dataset to be fair to the competition ?", "1546323": "100% agree" }, "source": "meta" }