![\"\"](\"https://i.imgur.com/itaCWjc.png\")
\nAt least those accounts were created with 2 days difference lol
I do not know if it is just me, but over the last few days I get the feeling there has suddenly appeared a number of recently created \"grey goose\" novice accounts, with absolutely no prior activity on kaggle of any kind, but whose submissions are all in the medals section of the leaderboard…
\n(Update: It seems that kaggle removed 10 teams form the medals section of the leaderboard)
", "messages": [ { "id": "1436681", "postDate": "08/03/2021 15:17:18", "content": "I do not know if it is just me, but over the last few days I get the feeling there has suddenly appeared a number of recently created \"grey goose\" novice accounts, with absolutely no prior activity on kaggle of any kind, but whose submissions are all in the medals section of the leaderboard…
\n(Update: It seems that kaggle removed 10 teams form the medals section of the leaderboard)
", "rawMarkdown": "I do not know if it is just me, but over the last few days I get the feeling there has suddenly appeared a number of recently created \"grey goose\" novice accounts, with absolutely no prior activity on kaggle of any kind, but whose submissions are all in the medals section of the leaderboard...\n\n(**Update**: It seems that kaggle removed 10 teams form the medals section of the leaderboard)", "votes": null }, { "id": "1439239", "postDate": "08/03/2021 16:51:04", "content": "I can see a few of those too, there's even one with 1 submission and a 16 days old account that is on the medal zone.
\nI believe that kaggle will end up banning those accounts, but i wonder if there's any process to spot the main account that is using this to probe the LB or they will just get away with the free info
", "rawMarkdown": "I can see a few of those too, there's even one with 1 submission and a 16 days old account that is on the medal zone.\n\nI believe that kaggle will end up banning those accounts, but i wonder if there's any process to spot the main account that is using this to probe the LB or they will just get away with the free info", "votes": null }, { "id": "1439418", "postDate": "08/03/2021 16:59:33", "content": "…It was suggested some time ago by someone that there actually exists a black market for kaggle accounts that have medals associated with them…
", "rawMarkdown": "...It was suggested some time ago by someone that there actually exists a black market for kaggle accounts that have medals associated with them...", "votes": null }, { "id": "1439613", "postDate": "08/03/2021 17:11:36", "content": "It makes perfect sense for this to exist, kaggle accounts can easily be one of the best evidences of true DS competency, i'm working my brains out trying to get some ranks here so i can use this as a differential.
\nWe should have a competition designed to find those cheaters and their main accounts 😂
", "rawMarkdown": "It makes perfect sense for this to exist, kaggle accounts can easily be one of the best evidences of true DS competency, i'm working my brains out trying to get some ranks here so i can use this as a differential.\n\nWe should have a competition designed to find those cheaters and their main accounts 😂", "votes": null }, { "id": "1440314", "postDate": "08/03/2021 18:08:45", "content": "
\nAt least those accounts were created with 2 days difference lol
![\"\"](\"https://i.imgur.com/7lQbkLk.png\")
\n¯¯_(ヅ)_/¯
Most of them were created 6~7 days ago, how fortunate they are to create their accounts on the rules acceptance deadline? And even manage to get such a score in a week.
\nVery likely that they are fine tuning their parameters like that, and those are the ones who didn't even bothered to mess up some of the data to create a low score baseline before tuning it, imagine how many ppl are actually doing this.
\nThe best i can do is work on my solution and hope this don't affect me much, but its kind of upsetting
", "rawMarkdown": "Most of them were created 6~7 days ago, how fortunate they are to create their accounts on the rules acceptance deadline? And even manage to get such a score in a week.\n\nVery likely that they are fine tuning their parameters like that, and those are the ones who didn't even bothered to mess up some of the data to create a low score baseline before tuning it, imagine how many ppl are actually doing this.\n\nThe best i can do is work on my solution and hope this don't affect me much, but its kind of upsetting", "votes": null }, { "id": "1449190", "postDate": "08/04/2021 20:25:28", "content": "Oh…
\n
![\"\"](\"https://i.imgur.com/wXzpN1m.png\")
\n4/5 accounts have been created 11 days ago. Seems like a significant part of the leaderboard should be cleansed at this point, but I dont really know if that would happen.
Tired of seeing those cheaters and shamed of them. Does anyone know how long the kaggle staff will finish the verification of the final score? I am really looking forward to see what is our real rank. Our team was shake down, sad☹️
", "rawMarkdown": "Tired of seeing those cheaters and shamed of them. Does anyone know how long the kaggle staff will finish the verification of the final score? I am really looking forward to see what is our real rank. Our team was shake down, sad☹️", "votes": null }, { "id": "1449824", "postDate": "08/05/2021 01:05:02", "content": "It usually takes a day, but this is the first time i see so many suspect accounts on the medal zone, not sure what is going to happen, but we should get some sort of official announcement about that.
\nI'm not very worried about the grey goose accounts, they should be deleted soon, but i really wish kaggle has a way to spot their main accounts and punish them somehow
\nIn any case, cheer up mate, we got our medals 😄
", "rawMarkdown": "It usually takes a day, but this is the first time i see so many suspect accounts on the medal zone, not sure what is going to happen, but we should get some sort of official announcement about that.\n\nI'm not very worried about the grey goose accounts, they should be deleted soon, but i really wish kaggle has a way to spot their main accounts and punish them somehow\n\nIn any case, cheer up mate, we got our medals 😄", "votes": null }, { "id": "1449839", "postDate": "08/05/2021 01:16:42", "content": "Congratulations for your medals ! There are some of us which those fake grey goose accounts pushed them outside of the medal zones. From my count, there are at least 3-4 really suspect ones in the medal zone (created several days before the competition ended). I am curious to see how Kaggle will handle them.
", "rawMarkdown": "Congratulations for your medals ! There are some of us which those fake grey goose accounts pushed them outside of the medal zones. From my count, there are at least 3-4 really suspect ones in the medal zone (created several days before the competition ended). I am curious to see how Kaggle will handle them.", "votes": null }, { "id": "1449857", "postDate": "08/05/2021 01:31:33", "content": "Thanks, my friend. I got the same worry as yours. Hope those cheaters' main accounts can be detected as well, but I guess it is quite hard. Let's see what is going on in the coming days.
\nAnyway, I learned a lot from this game. Congrats, we did enter the model zone, victorasso. 😄
Yes, let's waiting for the formal replying.🤝
", "rawMarkdown": "Yes, let's waiting for the formal replying.🤝", "votes": null }, { "id": "1450131", "postDate": "08/05/2021 03:47:29", "content": "Dear @vladvdv
\nI see you are currently 101st, so these dubious \"competitors\" must really sting at the moment. I sincerely hope the bad actors are weeded out over the next few days and you (as well as everyone else on the leaderboard) get the position that you deserve.
\nBTW: At the moment there are 822 teams on the leaderboard. It will be interesting to see how many remain after the results are verified…
\nAll the best,
\ncarl
Dear @avtobusbratiev
\nIn the end user dinglinglingling shook-down an astounding 706 places(!). I am not even sure how that was possible, unless they didn't notice that there were entire routes that were neither in the training data, nor in the public leaderboard evaluation (a bit sneaky it must be said).
\nAll the best,
\ncarl
So on point.
\nIt's time dummy accounts are checked on Kaggle.
Dear @kritidoneria
\nIndeed. However, one must at all costs avoid false positives; the vast majority of novice accounts are totally legitimate, and there are amazing data scientists on kaggle that have not fully registered yet or added a profile image, but are more than capable of winning competitions. The fraudulent accounts only really reveal themselves by \"implausible\" acts, such as joining kaggle in the last days of the competition, then making one or two submissions that go straight to the top of the leadeboard. Or pairs of accounts that submit exactly the same score almost simultaneously, which maybe are being used to get around the daily submission limit.
\nThat said, the kaggle team have a lot of experience dealing with the situation, and hopefully their impact will be minimal.
\nAll the best,
\ncarl
Thank you @carlmcbrideellis
\nThe Kaggle community needs to feel safe from these practices, otherwise nothing will stop more and more competitors to go on this road, why to have 5 submissions per day when they can have 10, or 20, or 50 ? Or more GPU time, or why choose 2 best submission for the private dataset when you can choose 4, or 12 with multiple accounts.
\nIn the same time, we should realize that this is not an easy thing to do. The main challenge for the Kaggle team is how to prove an account is a fake one ? ok, if it was created with 2 days before the deadline, it has 2 submissions and it is on the 7 place it is clearly dubious, but how can this be proven ?
Ip checks and automated code plagiarism detection can definitely help with that. Of course kaggle won't be able to totally solve this issue, but its possible to make it very hard to bypass
", "rawMarkdown": "Ip checks and automated code plagiarism detection can definitely help with that. Of course kaggle won't be able to totally solve this issue, but its possible to make it very hard to bypass", "votes": null }, { "id": "1452733", "postDate": "08/05/2021 17:19:35", "content": "Btw @vladvdv i hope you get this medal, being out bcz of that should be very upsetting
", "rawMarkdown": "Btw @vladvdv i hope you get this medal, being out bcz of that should be very upsetting", "votes": null }, { "id": "1452812", "postDate": "08/05/2021 17:36:41", "content": "Thank you for the kind words @victorasso . To be honest it is upsetting investing time and resources and then see lots of newly created accounts joining with 2 days ahead having 2-3 submissions and being in top 30.
\nBut, it is what it is, I try to take this as a challenge and perform better next time. In the end, what matters is the knowledge we get from each competition, knowledge that makes us better.
It seems you got it 😄
\nCongratz
", "rawMarkdown": "It seems you got it 😄\n\nCongratz", "votes": null }, { "id": "1452832", "postDate": "08/05/2021 17:43:27", "content": "Yes, just saw it ! it seems that 10 teams from top 100 were eliminated
", "rawMarkdown": "Yes, just saw it ! it seems that 10 teams from top 100 were eliminated", "votes": null }, { "id": "1453508", "postDate": "08/05/2021 22:14:25", "content": "From what I can see 12 teams have been excluded.
", "rawMarkdown": "From what I can see 12 teams have been excluded.", "votes": null }, { "id": "1453584", "postDate": "08/05/2021 23:31:13", "content": "Most likely this was a dummy account that was using his pipeline only on the public data, it should be pretty easy to do just following what people posted here
\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/250977
\nThis way someone who is not a grey goose got some advantage on top of the rest of us, it seems that only dummy accounts were excluded, so people actually got away with this technique.
\nIf you did this and are reading this, shame on you, i hope one day kaggle implement something that will find you and remove your ranks
", "rawMarkdown": "Most likely this was a dummy account that was using his pipeline only on the public data, it should be pretty easy to do just following what people posted here\n\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/250977\n\nThis way someone who is not a grey goose got some advantage on top of the rest of us, it seems that only dummy accounts were excluded, so people actually got away with this technique.\n\nIf you did this and are reading this, shame on you, i hope one day kaggle implement something that will find you and remove your ranks", "votes": null }, { "id": "1454498", "postDate": "08/06/2021 08:05:26", "content": "Hi, our team is new to kaggle. We major in gnss positioning and knew this competition from ION GNSS 2020. And I notice that there are teams that publish solution in Discussion. But I dont find any requirement about it. Is it necessary for us to publish solution or it is required on other Kaggle competition? Thx!
", "rawMarkdown": "Hi, our team is new to kaggle. We major in gnss positioning and knew this competition from ION GNSS 2020. And I notice that there are teams that publish solution in Discussion. But I dont find any requirement about it. Is it necessary for us to publish solution or it is required on other Kaggle competition? Thx!", "votes": null }, { "id": "1454513", "postDate": "08/06/2021 08:12:35", "content": "Dear @ivwvvy
\nNo, there is absolutely no requirement at all, it is just nice to see how the very top people, like yourself and your team, approached the problem. It is also a great way to learn from the best, and perhaps try out their ideas in the next competition!
\nWell done and all the best,
\ncarl
Got it. Thx a lot! Some of the ideas are very interesting and effective!
", "rawMarkdown": "Got it. Thx a lot! Some of the ideas are very interesting and effective!", "votes": null }, { "id": "1456257", "postDate": "08/06/2021 20:46:20", "content": "I like that approach @carlmcbrideellis
", "rawMarkdown": "I like that approach @carlmcbrideellis", "votes": null } ], "comments": [ { "id": 1439239, "author_name": "victorasso", "author_url": "", "post_date": "08/03/2021 16:51:04", "content": "I can see a few of those too, there's even one with 1 submission and a 16 days old account that is on the medal zone.
\nI believe that kaggle will end up banning those accounts, but i wonder if there's any process to spot the main account that is using this to probe the LB or they will just get away with the free info
", "votes": null, "replies": [ { "id": 1439418, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "08/03/2021 16:59:33", "content": "…It was suggested some time ago by someone that there actually exists a black market for kaggle accounts that have medals associated with them…
", "votes": null, "replies": [] }, { "id": 1439613, "author_name": "victorasso", "author_url": "", "post_date": "08/03/2021 17:11:36", "content": "It makes perfect sense for this to exist, kaggle accounts can easily be one of the best evidences of true DS competency, i'm working my brains out trying to get some ranks here so i can use this as a differential.
\nWe should have a competition designed to find those cheaters and their main accounts 😂
", "votes": null, "replies": [] } ] }, { "id": 1440314, "author_name": "avtobusbratiev", "author_url": "", "post_date": "08/03/2021 18:08:45", "content": "
\nAt least those accounts were created with 2 days difference lol
\n¯¯_(ヅ)_/¯
Most of them were created 6~7 days ago, how fortunate they are to create their accounts on the rules acceptance deadline? And even manage to get such a score in a week.
\nVery likely that they are fine tuning their parameters like that, and those are the ones who didn't even bothered to mess up some of the data to create a low score baseline before tuning it, imagine how many ppl are actually doing this.
\nThe best i can do is work on my solution and hope this don't affect me much, but its kind of upsetting
", "votes": null, "replies": [] }, { "id": 1450252, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "08/05/2021 04:24:00", "content": "Dear @avtobusbratiev
\nIn the end user dinglinglingling shook-down an astounding 706 places(!). I am not even sure how that was possible, unless they didn't notice that there were entire routes that were neither in the training data, nor in the public leaderboard evaluation (a bit sneaky it must be said).
\nAll the best,
\ncarl
Most likely this was a dummy account that was using his pipeline only on the public data, it should be pretty easy to do just following what people posted here
\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/250977
\nThis way someone who is not a grey goose got some advantage on top of the rest of us, it seems that only dummy accounts were excluded, so people actually got away with this technique.
\nIf you did this and are reading this, shame on you, i hope one day kaggle implement something that will find you and remove your ranks
", "votes": null, "replies": [] } ] }, { "id": 1449190, "author_name": "t88take", "author_url": "", "post_date": "08/04/2021 20:25:28", "content": "Oh…
\n
\n4/5 accounts have been created 11 days ago. Seems like a significant part of the leaderboard should be cleansed at this point, but I dont really know if that would happen.
Tired of seeing those cheaters and shamed of them. Does anyone know how long the kaggle staff will finish the verification of the final score? I am really looking forward to see what is our real rank. Our team was shake down, sad☹️
", "votes": null, "replies": [ { "id": 1449824, "author_name": "victorasso", "author_url": "", "post_date": "08/05/2021 01:05:02", "content": "It usually takes a day, but this is the first time i see so many suspect accounts on the medal zone, not sure what is going to happen, but we should get some sort of official announcement about that.
\nI'm not very worried about the grey goose accounts, they should be deleted soon, but i really wish kaggle has a way to spot their main accounts and punish them somehow
\nIn any case, cheer up mate, we got our medals 😄
", "votes": null, "replies": [] }, { "id": 1449839, "author_name": "vladvdv", "author_url": "", "post_date": "08/05/2021 01:16:42", "content": "Congratulations for your medals ! There are some of us which those fake grey goose accounts pushed them outside of the medal zones. From my count, there are at least 3-4 really suspect ones in the medal zone (created several days before the competition ended). I am curious to see how Kaggle will handle them.
", "votes": null, "replies": [] }, { "id": 1449857, "author_name": "alvinai9603", "author_url": "", "post_date": "08/05/2021 01:31:33", "content": "Thanks, my friend. I got the same worry as yours. Hope those cheaters' main accounts can be detected as well, but I guess it is quite hard. Let's see what is going on in the coming days.
\nAnyway, I learned a lot from this game. Congrats, we did enter the model zone, victorasso. 😄
Yes, let's waiting for the formal replying.🤝
", "votes": null, "replies": [] }, { "id": 1450131, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "08/05/2021 03:47:29", "content": "Dear @vladvdv
\nI see you are currently 101st, so these dubious \"competitors\" must really sting at the moment. I sincerely hope the bad actors are weeded out over the next few days and you (as well as everyone else on the leaderboard) get the position that you deserve.
\nBTW: At the moment there are 822 teams on the leaderboard. It will be interesting to see how many remain after the results are verified…
\nAll the best,
\ncarl
Thank you @carlmcbrideellis
\nThe Kaggle community needs to feel safe from these practices, otherwise nothing will stop more and more competitors to go on this road, why to have 5 submissions per day when they can have 10, or 20, or 50 ? Or more GPU time, or why choose 2 best submission for the private dataset when you can choose 4, or 12 with multiple accounts.
\nIn the same time, we should realize that this is not an easy thing to do. The main challenge for the Kaggle team is how to prove an account is a fake one ? ok, if it was created with 2 days before the deadline, it has 2 submissions and it is on the 7 place it is clearly dubious, but how can this be proven ?
Ip checks and automated code plagiarism detection can definitely help with that. Of course kaggle won't be able to totally solve this issue, but its possible to make it very hard to bypass
", "votes": null, "replies": [] }, { "id": 1452733, "author_name": "victorasso", "author_url": "", "post_date": "08/05/2021 17:19:35", "content": "Btw @vladvdv i hope you get this medal, being out bcz of that should be very upsetting
", "votes": null, "replies": [] }, { "id": 1452812, "author_name": "vladvdv", "author_url": "", "post_date": "08/05/2021 17:36:41", "content": "Thank you for the kind words @victorasso . To be honest it is upsetting investing time and resources and then see lots of newly created accounts joining with 2 days ahead having 2-3 submissions and being in top 30.
\nBut, it is what it is, I try to take this as a challenge and perform better next time. In the end, what matters is the knowledge we get from each competition, knowledge that makes us better.
It seems you got it 😄
\nCongratz
", "votes": null, "replies": [] }, { "id": 1452832, "author_name": "vladvdv", "author_url": "", "post_date": "08/05/2021 17:43:27", "content": "Yes, just saw it ! it seems that 10 teams from top 100 were eliminated
", "votes": null, "replies": [] }, { "id": 1453508, "author_name": "avtobusbratiev", "author_url": "", "post_date": "08/05/2021 22:14:25", "content": "From what I can see 12 teams have been excluded.
", "votes": null, "replies": [] } ] }, { "id": 1451165, "author_name": "kritidoneria", "author_url": "", "post_date": "08/05/2021 09:19:42", "content": "So on point.
\nIt's time dummy accounts are checked on Kaggle.
Dear @kritidoneria
\nIndeed. However, one must at all costs avoid false positives; the vast majority of novice accounts are totally legitimate, and there are amazing data scientists on kaggle that have not fully registered yet or added a profile image, but are more than capable of winning competitions. The fraudulent accounts only really reveal themselves by \"implausible\" acts, such as joining kaggle in the last days of the competition, then making one or two submissions that go straight to the top of the leadeboard. Or pairs of accounts that submit exactly the same score almost simultaneously, which maybe are being used to get around the daily submission limit.
\nThat said, the kaggle team have a lot of experience dealing with the situation, and hopefully their impact will be minimal.
\nAll the best,
\ncarl
I like that approach @carlmcbrideellis
", "votes": null, "replies": [] } ] }, { "id": 1454498, "author_name": "ivwvvy", "author_url": "", "post_date": "08/06/2021 08:05:26", "content": "Hi, our team is new to kaggle. We major in gnss positioning and knew this competition from ION GNSS 2020. And I notice that there are teams that publish solution in Discussion. But I dont find any requirement about it. Is it necessary for us to publish solution or it is required on other Kaggle competition? Thx!
", "votes": null, "replies": [ { "id": 1454513, "author_name": "carlmcbrideellis", "author_url": "", "post_date": "08/06/2021 08:12:35", "content": "Dear @ivwvvy
\nNo, there is absolutely no requirement at all, it is just nice to see how the very top people, like yourself and your team, approached the problem. It is also a great way to learn from the best, and perhaps try out their ideas in the next competition!
\nWell done and all the best,
\ncarl
Got it. Thx a lot! Some of the ideas are very interesting and effective!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1436681": "I do not know if it is just me, but over the last few days I get the feeling there has suddenly appeared a number of recently created \"grey goose\" novice accounts, with absolutely no prior activity on kaggle of any kind, but whose submissions are all in the medals section of the leaderboard...\n\n(**Update**: It seems that kaggle removed 10 teams form the medals section of the leaderboard)", "1439239": "I can see a few of those too, there's even one with 1 submission and a 16 days old account that is on the medal zone.\n\nI believe that kaggle will end up banning those accounts, but i wonder if there's any process to spot the main account that is using this to probe the LB or they will just get away with the free info", "1439418": "...It was suggested some time ago by someone that there actually exists a black market for kaggle accounts that have medals associated with them...", "1439613": "It makes perfect sense for this to exist, kaggle accounts can easily be one of the best evidences of true DS competency, i'm working my brains out trying to get some ranks here so i can use this as a differential.\n\nWe should have a competition designed to find those cheaters and their main accounts 😂", "1440314": "\nAt least those accounts were created with 2 days difference lol", "1441951": "\n¯¯\\_(ヅ)_/¯", "1442231": "Most of them were created 6~7 days ago, how fortunate they are to create their accounts on the rules acceptance deadline? And even manage to get such a score in a week.\n\nVery likely that they are fine tuning their parameters like that, and those are the ones who didn't even bothered to mess up some of the data to create a low score baseline before tuning it, imagine how many ppl are actually doing this.\n\nThe best i can do is work on my solution and hope this don't affect me much, but its kind of upsetting", "1449190": "Oh...\n", "1449218": "\n4/5 accounts have been created 11 days ago. Seems like ~~1/3~~ a significant part of the leaderboard should be cleansed at this point, but I dont really know if that would happen.", "1449765": "Tired of seeing those cheaters and shamed of them. Does anyone know how long the kaggle staff will finish the verification of the final score? I am really looking forward to see what is our real rank. Our team was shake down, sad☹️", "1449824": "It usually takes a day, but this is the first time i see so many suspect accounts on the medal zone, not sure what is going to happen, but we should get some sort of official announcement about that.\n\nI'm not very worried about the grey goose accounts, they should be deleted soon, but i really wish kaggle has a way to spot their main accounts and punish them somehow\n\nIn any case, cheer up mate, we got our medals 😄", "1449839": "Congratulations for your medals ! There are some of us which those fake grey goose accounts pushed them outside of the medal zones. From my count, there are at least 3-4 really suspect ones in the medal zone (created several days before the competition ended). I am curious to see how Kaggle will handle them.", "1449857": "Thanks, my friend. I got the same worry as yours. Hope those cheaters' main accounts can be detected as well, but I guess it is quite hard. Let's see what is going on in the coming days.\nAnyway, I learned a lot from this game. Congrats, we did enter the model zone, victorasso. 😄", "1449858": "Yes, let's waiting for the formal replying.🤝", "1450131": "Dear @vladvdv \n\nI see you are currently 101st, so these dubious \"competitors\" must really sting at the moment. I sincerely hope the bad actors are weeded out over the next few days and you (as well as everyone else on the leaderboard) get the position that you deserve.\n\nBTW: At the moment there are 822 teams on the leaderboard. It will be interesting to see how many remain after the results are verified...\n\nAll the best,\ncarl", "1450252": "Dear @avtobusbratiev \n\nIn the end user **dinglinglingling** shook-down an astounding 706 places(!). I am not even sure how that was possible, unless they didn't notice that there were entire routes that were neither in the training data, nor in the public leaderboard evaluation (a bit sneaky it must be said).\n\nAll the best,\ncarl", "1451165": "So on point.\nIt's time dummy accounts are checked on Kaggle.", "1451224": "Dear @kritidoneria \n\nIndeed. However, one must **at all costs** avoid false positives; the vast majority of novice accounts are totally legitimate, and there are amazing data scientists on kaggle that have not fully registered yet or added a profile image, but are more than capable of winning competitions. The fraudulent accounts only really reveal themselves by \"implausible\" acts, such as joining kaggle in the last days of the competition, then making one or two submissions that go straight to the top of the leadeboard. Or pairs of accounts that submit exactly the same score almost simultaneously, which maybe are being used to get around the daily submission limit.\n\nThat said, the kaggle team have a lot of experience dealing with the situation, and hopefully their impact will be minimal.\n\nAll the best,\ncarl", "1452272": "Thank you @carlmcbrideellis \nThe Kaggle community needs to feel safe from these practices, otherwise nothing will stop more and more competitors to go on this road, why to have 5 submissions per day when they can have 10, or 20, or 50 ? Or more GPU time, or why choose 2 best submission for the private dataset when you can choose 4, or 12 with multiple accounts.\nIn the same time, we should realize that this is not an easy thing to do. The main challenge for the Kaggle team is how to prove an account is a fake one ? ok, if it was created with 2 days before the deadline, it has 2 submissions and it is on the 7 place it is clearly dubious, but how can this be proven ?", "1452300": "~~logistic regression~~", "1452722": "Ip checks and automated code plagiarism detection can definitely help with that. Of course kaggle won't be able to totally solve this issue, but its possible to make it very hard to bypass", "1452733": "Btw @vladvdv i hope you get this medal, being out bcz of that should be very upsetting", "1452812": "Thank you for the kind words @victorasso . To be honest it is upsetting investing time and resources and then see lots of newly created accounts joining with 2 days ahead having 2-3 submissions and being in top 30. \nBut, it is what it is, I try to take this as a challenge and perform better next time. In the end, what matters is the knowledge we get from each competition, knowledge that makes us better.", "1452827": "It seems you got it 😄\n\nCongratz", "1452832": "Yes, just saw it ! it seems that 10 teams from top 100 were eliminated", "1453508": "From what I can see 12 teams have been excluded.", "1453584": "Most likely this was a dummy account that was using his pipeline only on the public data, it should be pretty easy to do just following what people posted here\n\nhttps://www.kaggle.com/c/google-smartphone-decimeter-challenge/discussion/250977\n\nThis way someone who is not a grey goose got some advantage on top of the rest of us, it seems that only dummy accounts were excluded, so people actually got away with this technique.\n\nIf you did this and are reading this, shame on you, i hope one day kaggle implement something that will find you and remove your ranks", "1454498": "Hi, our team is new to kaggle. We major in gnss positioning and knew this competition from ION GNSS 2020. And I notice that there are teams that publish solution in Discussion. But I dont find any requirement about it. Is it necessary for us to publish solution or it is required on other Kaggle competition? Thx!", "1454513": "Dear @ivwvvy \n\nNo, there is absolutely no requirement at all, it is just nice to see how the very top people, like yourself and your team, approached the problem. It is also a great way to learn from the best, and perhaps try out their ideas in the next competition!\n\nWell done and all the best,\ncarl", "1454533": "Got it. Thx a lot! Some of the ideas are very interesting and effective!", "1456257": "I like that approach @carlmcbrideellis" }, "source": "meta" }![\"\"](\"https://pbs.twimg.com/media/E7-RxG_VoAMBZj2?format=jpg&name=large\")