{ "id": 381360, "title": "One more thread about cheaters", "url": "/competitions/otto-recommender-system/discussion/381360", "author_name": "Ilia los", "post_date": "2023-01-26T11:50:32.955000", "votes": 84, "comment_count": 47, "views": 0, "content": "

You've already read about the situation that affected two teams on top of the leaderboard. Unfortunately my team was involved as well.

\n

You can read these threads for more details:
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381321

\n

I would like to cover two main things:
\n1) explain one more time what happened, but this time from my and @dempton side
\n2) explain a way how we can reliably prove that these guys did share the code between each other during the competition

\n

So what happened?
\nAt the last day before the merge deadline we were contacted by @runningz with a proposition to merge. His team's score was 601 at the moment so we rejected it, because strongly believe that could achieve such score very soon. Later that day he made a submission with a score 603 and after that we decided to accept the invitation. After the merge in a private slack space he told us that the best single model has score 599, but there are no submission with such score. We found it suspicion and decided to have a look on his history of participations. This is not an evidence, but just check that almost every time when @runningz has a high place, @poteman is very close.

\n

It became even more suspicion when on the next day @bruceqdu told us that the best single model score is 597 and @runningz agreed by telling that mentioning 599 yesterday was a mistake. After that we decided to force @runningz to prove the way how he got that 603 result. He sent us several screenshots with his algorithm to ensemble the submissions and told that he cannot reproduce the 603 submission anymore.
\n\"\"
\nHave a look that images were deleted by him after that. Later we will find out why he did that.

\n

We continued to push him, because after that it was much clearer that something isn't right. And we decided to go even more deeper. Of course on that particular moment we didn't know for sure that @runningz and @poteman are cheaters, but the bait worked perfectly.

\n

\"\"

\n

After that he started to clear slack history with details about his ensembling techniques. But we have saved all images and text messages before that, because expected something like that from him. And in general his behaviour was the answer for us. After that he hasn't sent any messages yet and he hasn't answer us anymore.

\n

So what I would like to propose to try prove that these two guys are cheaters.
\nAfter the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603. We have his code for ensembling (I'm going to share it here after the end of the competition), we just need to get the correct list of submission files to stack with the right coefficients. And I'm pretty sure that the list will have at least one submission outside of our team. I think if we get the exact match for all 5015409 lines then it will be absolutely clear that it is not a coincided.

\n

@mhyodo I hope that you will be ready to collaborate on this after the end of the competition
\n@pnormann please confirm that the way that I described above will lead us to ban these two gentlemen forever? Of course in case if we get exactly the same submission file. I hope that you take it all serious, because @poteman is currently top20 in kaggle competitions overall rating and @runningz is one gold medal away from competitions GM title. Cheating is a bad thing at all levels, but when such high ranked persons are involved then I think it's much worse.

\n

And about our participation.
\nWe decided not to continue, because @runningz is holding team lead position, we've already spent 3 days on that staff and at the end we are just too demoralised. I hope kaggle will find a way to preserve us in this competitions in a some honest way. For example, to remove @runnigz and use 2 best submissions that we sent with @dempton before the merge as our team's final submissions.

", "messages": [ { "id": 2116296, "postDate": "2023-01-26T11:50:32.957Z", "content": "

You've already read about the situation that affected two teams on top of the leaderboard. Unfortunately my team was involved as well.

\n

You can read these threads for more details:
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318
\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381321

\n

I would like to cover two main things:
\n1) explain one more time what happened, but this time from my and @dempton side
\n2) explain a way how we can reliably prove that these guys did share the code between each other during the competition

\n

So what happened?
\nAt the last day before the merge deadline we were contacted by @runningz with a proposition to merge. His team's score was 601 at the moment so we rejected it, because strongly believe that could achieve such score very soon. Later that day he made a submission with a score 603 and after that we decided to accept the invitation. After the merge in a private slack space he told us that the best single model has score 599, but there are no submission with such score. We found it suspicion and decided to have a look on his history of participations. This is not an evidence, but just check that almost every time when @runningz has a high place, @poteman is very close.

\n

It became even more suspicion when on the next day @bruceqdu told us that the best single model score is 597 and @runningz agreed by telling that mentioning 599 yesterday was a mistake. After that we decided to force @runningz to prove the way how he got that 603 result. He sent us several screenshots with his algorithm to ensemble the submissions and told that he cannot reproduce the 603 submission anymore.
\n\"\"
\nHave a look that images were deleted by him after that. Later we will find out why he did that.

\n

We continued to push him, because after that it was much clearer that something isn't right. And we decided to go even more deeper. Of course on that particular moment we didn't know for sure that @runningz and @poteman are cheaters, but the bait worked perfectly.

\n

\"\"

\n

After that he started to clear slack history with details about his ensembling techniques. But we have saved all images and text messages before that, because expected something like that from him. And in general his behaviour was the answer for us. After that he hasn't sent any messages yet and he hasn't answer us anymore.

\n

So what I would like to propose to try prove that these two guys are cheaters.
\nAfter the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603. We have his code for ensembling (I'm going to share it here after the end of the competition), we just need to get the correct list of submission files to stack with the right coefficients. And I'm pretty sure that the list will have at least one submission outside of our team. I think if we get the exact match for all 5015409 lines then it will be absolutely clear that it is not a coincided.

\n

@mhyodo I hope that you will be ready to collaborate on this after the end of the competition
\n@pnormann please confirm that the way that I described above will lead us to ban these two gentlemen forever? Of course in case if we get exactly the same submission file. I hope that you take it all serious, because @poteman is currently top20 in kaggle competitions overall rating and @runningz is one gold medal away from competitions GM title. Cheating is a bad thing at all levels, but when such high ranked persons are involved then I think it's much worse.

\n

And about our participation.
\nWe decided not to continue, because @runningz is holding team lead position, we've already spent 3 days on that staff and at the end we are just too demoralised. I hope kaggle will find a way to preserve us in this competitions in a some honest way. For example, to remove @runnigz and use 2 best submissions that we sent with @dempton before the merge as our team's final submissions.

", "rawMarkdown": "You've already read about the situation that affected two teams on top of the leaderboard. Unfortunately my team was involved as well.\n\nYou can read these threads for more details:\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381321\n\nI would like to cover two main things:\n1) explain one more time what happened, but this time from my and @dempton side\n2) explain a way how we can reliably prove that these guys did share the code between each other during the competition\n\n**So what happened?**\nAt the last day before the merge deadline we were contacted by @runningz with a proposition to merge. His team's score was 601 at the moment so we rejected it, because strongly believe that could achieve such score very soon. Later that day he made a submission with a score 603 and after that we decided to accept the invitation. After the merge in a private slack space he told us that the best single model has score 599, but there are no submission with such score. We found it suspicion and decided to have a look on his history of participations. **This is not an evidence**, but just check that almost every time when @runningz has a high place, @poteman is very close. \n\nIt became even more suspicion when on the next day @bruceqdu told us that the best single model score is 597 and @runningz agreed by telling that mentioning 599 yesterday was a mistake. After that we decided to force @runningz to prove the way how he got that 603 result. He sent us several screenshots with his algorithm to ensemble the submissions and told that he cannot reproduce the 603 submission anymore.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1082778%2Fd5739c3d0a6310bcda9a2f4907878eb1%2FScreenshot%202023-01-26%20at%2010.59.29.png?generation=1674730469831845&alt=media)\nHave a look that images were deleted by him after that. Later we will find out why he did that.\n\nWe continued to push him, because after that it was much clearer that something isn't right. And we decided to go even more deeper. Of course on that particular moment we didn't know for sure that @runningz and @poteman are cheaters, but the bait worked perfectly.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1082778%2F04c09ab93b99c46dec8861330185fa2f%2FScreenshot%202023-01-26%20at%2011.57.37.png?generation=1674730672995028&alt=media)\n\nAfter that he started to clear slack history with details about his ensembling techniques. But we have saved all images and text messages before that, because expected something like that from him. And in general his behaviour was the answer for us. After that he hasn't sent any messages yet and he hasn't answer us anymore.\n\n**So what I would like to propose to try prove that these two guys are cheaters.**\nAfter the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603. We have his code for ensembling (I'm going to share it here after the end of the competition), we just need to get the correct list of submission files to stack with the right coefficients. And I'm pretty sure that the list will have at least one submission outside of our team. I think if we get the exact match for all 5015409 lines then it will be absolutely clear that it is not a coincided.\n\n@mhyodo I hope that you will be ready to collaborate on this after the end of the competition\n@pnormann please confirm that the way that I described above will lead us to ban these two gentlemen forever? Of course in case if we get exactly the same submission file. I hope that you take it all serious, because @poteman is currently top20 in kaggle competitions overall rating and @runningz is one gold medal away from competitions GM title. Cheating is a bad thing at all levels, but when such high ranked persons are involved then I think it's much worse.\n\n**And about our participation.**\nWe decided not to continue, because @runningz is holding team lead position, we've already spent 3 days on that staff and at the end we are just too demoralised. I hope kaggle will find a way to preserve us in this competitions in a some honest way. For example, to remove @runnigz and use 2 best submissions that we sent with @dempton before the merge as our team's final submissions.", "votes": 84 }, { "id": 2119821, "postDate": "2023-01-29T06:08:22.837Z", "content": "

@wcukierski I hope kaggle team deal with cheaters strictly, for the fairness and the honor of all faithful kagglers.

", "rawMarkdown": "@wcukierski I hope kaggle team deal with cheaters strictly, for the fairness and the honor of all faithful kagglers.", "votes": 15 }, { "id": 2117459, "postDate": "2023-01-27T09:40:18.610Z", "content": "

People talk more and more about suspicious activity on Kaggle. A lot of people put hard and long hours into getting a high LB position. Often it feels like an extra job (especially when starting out).

\n

Kaggle has a very high reputation in the data science/ML world. I just hope Kaggle takes good steps to assure cheating in any way is forbidden, sanctioned and made hard. There is no perfect system, but reading things like this just decreases motivation for everyone who works hard.

", "rawMarkdown": "People talk more and more about suspicious activity on Kaggle. A lot of people put hard and long hours into getting a high LB position. Often it feels like an extra job (especially when starting out).\n\nKaggle has a very high reputation in the data science/ML world. I just hope Kaggle takes good steps to assure cheating in any way is forbidden, sanctioned and made hard. There is no perfect system, but reading things like this just decreases motivation for everyone who works hard.", "votes": 15 }, { "id": 2117030, "postDate": "2023-01-26T23:01:35.540Z", "content": "

My biggest question is do they really exist on this earth?

", "rawMarkdown": "My biggest question is `do they really exist on this earth?`", "votes": 11, "replies": [ { "id": 2120919, "postDate": "2023-01-29T22:50:28.523Z", "content": "

Haha, you made me laugh

", "rawMarkdown": "Haha, you made me laugh" } ] }, { "id": 2116323, "postDate": "2023-01-26T12:17:09.290Z", "content": "

@ilyalos

\n

Thanks for the info! I would very much like to cooperate with you.
\nWe have a suggestion as well.
\nWe have 3 different results posted by Poteman in our team's sub. Each has a score of 0.595 / 0.596 / 0.6.
\nThese results may also be an ensemble of results from other teams. (It would be a quicker story if he had used the raw other team's submissions…)
\nAfter the competition is over, let's cross-check the submissions made by the alleged members with each other!

\n

PS: Poteman's side did not give any codes for the ensemble.

", "rawMarkdown": "@ilyalos \n\nThanks for the info! I would very much like to cooperate with you.\nWe have a suggestion as well.\nWe have 3 different results posted by Poteman in our team's sub. Each has a score of 0.595 / 0.596 / 0.6.\nThese results may also be an ensemble of results from other teams. (It would be a quicker story if he had used the raw other team's submissions...)\nAfter the competition is over, let's cross-check the submissions made by the alleged members with each other!\n\nPS: Poteman's side did not give any codes for the ensemble.", "votes": 11, "replies": [ { "id": 2116325, "postDate": "2023-01-26T12:22:44.617Z", "content": "

Hi @mhyodo ,
\nThank you for so quick response. Yes, probably they might be related to one of submissions that was made in our team by @bruceqdu (scores are very similar, except 0.6). Let's try to investigate it after the end of the competition.

", "rawMarkdown": "Hi @mhyodo ,\nThank you for so quick response. Yes, probably they might be related to one of submissions that was made in our team by @bruceqdu (scores are very similar, except 0.6). Let's try to investigate it after the end of the competition.", "votes": 4 } ] }, { "id": 2116331, "postDate": "2023-01-26T12:27:24.057Z", "content": "

My guess would be that finding the right coefficients is not guaranteed

\n\n

However, I think that even without such info there is more than enough proof that they are cheaters.
\nMy fear is that there are more than 2 people collaborating, and that the others get away with it.

", "rawMarkdown": "My guess would be that finding the right coefficients is not guaranteed\n- The code you have may not be the one he used for the 0.603 sub\n- There may be other files involved than the ones poteman's team mates have access to.\n\nHowever, I think that even without such info there is more than enough proof that they are cheaters.\nMy fear is that there are more than 2 people collaborating, and that the others get away with it.", "votes": 5, "replies": [ { "id": 2116335, "postDate": "2023-01-26T12:29:58.500Z", "content": "

Would be interesting to study the following thing:

\n\n

Actually this could be an interesting analysis across many Kagglers.

", "rawMarkdown": "Would be interesting to study the following thing:\n\n- Access META Kaggle leaderboards of all competitions these two competitors participated\n- Calculate the average rank difference across all other competitors across competitions\n- Rank this list (+ statistical significance vs. the average distance of all pairs)\n\nActually this could be an interesting analysis across many Kagglers.", "votes": 22, "replies": [ { "id": 2116355, "postDate": "2023-01-26T12:43:31.090Z", "content": "

It's about time to launch Kaggle fraud detection competition

", "rawMarkdown": "It's about time to launch Kaggle fraud detection competition", "votes": 19 }, { "id": 2120255, "postDate": "2023-01-29T13:09:22.213Z", "content": "

This could be an interesting analysis

", "rawMarkdown": "This could be an interesting analysis", "votes": 1 } ] }, { "id": 2116348, "postDate": "2023-01-26T12:39:51.807Z", "content": "

I think the fact that he tried to clear the message history speaks a lot.

\n

His code is very simple and I'm almost sure that he used it to get 603 submission. Again, I do not want to share it until the end of the competition, but you will see.
\nAlso at one of the screenshots he shared a list of submission file names that he tried to stack. So it simplifies the things a bit. But you're right, if more than 2 persons were involved then proposed plan will not succeed.

", "rawMarkdown": "I think the fact that he tried to clear the message history speaks a lot.\n\nHis code is very simple and I'm almost sure that he used it to get 603 submission. Again, I do not want to share it until the end of the competition, but you will see.\nAlso at one of the screenshots he shared a list of submission file names that he tried to stack. So it simplifies the things a bit. But you're right, if more than 2 persons were involved then proposed plan will not succeed.", "votes": 4, "replies": [ { "id": 2116354, "postDate": "2023-01-26T12:43:13.547Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 2116375, "postDate": "2023-01-26T13:00:56.467Z", "content": "

yes if they can't reproduce their submissions for the initial score (which they definitely can't) a 3rd guy is surely involved, helping them with the submissions. And letting them go away, means they will continue this in the future competitions too

", "rawMarkdown": "yes if they can't reproduce their submissions for the initial score (which they definitely can't) a 3rd guy is surely involved, helping them with the submissions. And letting them go away, means they will continue this in the future competitions too", "replies": [ { "id": 2116439, "postDate": "2023-01-26T13:49:22.420Z", "content": "

I think it's a bit simpler. They can reproduce it, but they just do not want to do that, because in that case they must provide us a list of submissions that they used for stacking.

", "rawMarkdown": "I think it's a bit simpler. They can reproduce it, but they just do not want to do that, because in that case they must provide us a list of submissions that they used for stacking.", "votes": 1 }, { "id": 2116445, "postDate": "2023-01-26T13:52:32.720Z", "content": "

sorry for the confusion, I meant the code which they used to make the initial submissions. They can't reproduce the code they used to make subs before teaming up

", "rawMarkdown": "sorry for the confusion, I meant the code which they used to make the initial submissions. They can't reproduce the code they used to make subs before teaming up" } ] }, { "id": 2117149, "postDate": "2023-01-27T04:14:49.997Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2121683, "postDate": "2023-01-30T13:41:49.487Z", "content": "

As far as i understand none of those two shared their code, and it goes waay into other competitions. This opens up a possibility of having more people collaborating. Ill now tag some guys (that I personally saw on kaggle contributing) from other competitions, can you guys share, if poteman or runningz actually provided code in any competition you did together? @ren4yu @rohitsingh9990 @iafoss @anonamename

", "rawMarkdown": "As far as i understand none of those two shared their code, and it goes waay into other competitions. This opens up a possibility of having more people collaborating. Ill now tag some guys (that I personally saw on kaggle contributing) from other competitions, can you guys share, if poteman or runningz actually provided code in any competition you did together? @ren4yu @rohitsingh9990 @iafoss @anonamename ", "votes": 5, "replies": [ { "id": 2122415, "postDate": "2023-01-30T23:16:21.390Z", "content": "

Hi, @bakeryproducts my team achieved the first place in recent Feedback3 competition, poteman was part of our team. He requested me to merge with him earlier in this competition and since he was solo gold winner for feedback2 competition i agreed to merge. But after merging he got busy with his office work and wasn't able to contribute in that competition (this is what he told after we won the competition and there is no reason for us to suspect otherwise). Since he didn't contributed, never shared any code. This was my first time collaborating with him so, i can't comment on any other competition he participated but atleast for Feedback3 comp. he never shared any code with us.

", "rawMarkdown": "Hi, @bakeryproducts my team achieved the first place in recent Feedback3 competition, poteman was part of our team. He requested me to merge with him earlier in this competition and since he was solo gold winner for feedback2 competition i agreed to merge. But after merging he got busy with his office work and wasn't able to contribute in that competition (this is what he told after we won the competition and there is no reason for us to suspect otherwise). Since he didn't contributed, never shared any code. This was my first time collaborating with him so, i can't comment on any other competition he participated but atleast for Feedback3 comp. he never shared any code with us.", "votes": 9 }, { "id": 2122433, "postDate": "2023-01-30T23:44:38.210Z", "content": "

@rohitsingh9990 That's the 4th instance where he did nothing. I've heard every kagglers who teamed up with him won't team up with him anymore.
\nWe should share these information to avoid further casualties.

", "rawMarkdown": "@rohitsingh9990 That's the 4th instance where he did nothing. I've heard every kagglers who teamed up with him won't team up with him anymore.\nWe should share these information to avoid further casualties.", "votes": 3 }, { "id": 2122491, "postDate": "2023-01-31T01:24:27.293Z", "content": "

Hi, @bakeryproducts
\nI teamed up with him in G2Net2 competition. In this competition as well, he got busy with his office work. He did not contribute anything during competition period. After competition ended, he told me that he had tried stacking with Out-Of-Fold data I created. However, he did not share any code with me.

", "rawMarkdown": "Hi, @bakeryproducts\nI teamed up with him in [G2Net2](https://www.kaggle.com/competitions/g2net-detecting-continuous-gravitational-waves) competition. In this competition as well, he got busy with his office work. He did not contribute anything during competition period. After competition ended, he told me that he had tried stacking with Out-Of-Fold data I created. However, he did not share any code with me.", "votes": 4 }, { "id": 2122964, "postDate": "2023-01-31T08:35:15.067Z", "content": "

thank you @rohitsingh9990 (i mean for now there is at least some reason to suspect otherwise?) thank you @anonamename , huge info

", "rawMarkdown": "thank you @rohitsingh9990 (i mean for now there is at least some reason to suspect otherwise?) thank you @anonamename , huge info", "votes": 2 } ] } ] }, { "id": 2116308, "postDate": "2023-01-26T12:02:12.830Z", "content": "

Again thank you coming out honestly, this is truly remarkable to see such honesty.

", "rawMarkdown": "Again thank you coming out honestly, this is truly remarkable to see such honesty.", "votes": 5, "replies": [ { "id": 2116338, "postDate": "2023-01-26T12:32:56.317Z", "content": "

I can say for all four of us (myself, @dempton, @bruceqdu, @hansu1017) that being a part of something unfair is really uncomfortably for us and we just cannot accept it. Hope most of the kagglers have the same position on that except very few individuals.

", "rawMarkdown": "I can say for all four of us (myself, @dempton, @bruceqdu, @hansu1017) that being a part of something unfair is really uncomfortably for us and we just cannot accept it. Hope most of the kagglers have the same position on that except very few individuals.", "votes": 7, "replies": [ { "id": 2116376, "postDate": "2023-01-26T13:01:48.063Z", "content": "

yes I really hope it too

", "rawMarkdown": "yes I really hope it too" } ] } ] }, { "id": 2121760, "postDate": "2023-01-30T14:35:42.040Z", "content": "

Wow. Maybe you should contact kaggle staff if not done yet, and discuss with them what evidence you need to provide to confound the suspected cheaters. Discussing evidence in front of them can help them escape. I think the mail is compliance@kaggle.com

\n

This is a really sad story for you and your honest team mates, but i hope some good will come out of it thanks to your actions. Thanks for that.

\n

FYI, a colleague of mine also spotted this and found it suspicious: \"almost every time when @runningz has a high place, @poteman is very close. \"

", "rawMarkdown": "Wow. Maybe you should contact kaggle staff if not done yet, and discuss with them what evidence you need to provide to confound the suspected cheaters. Discussing evidence in front of them can help them escape. I think the mail is compliance@kaggle.com\n\nThis is a really sad story for you and your honest team mates, but i hope some good will come out of it thanks to your actions. Thanks for that.\n\nFYI, a colleague of mine also spotted this and found it suspicious: \"almost every time when @runningz has a high place, @poteman is very close. \"\n\n", "votes": 4, "replies": [ { "id": 2121765, "postDate": "2023-01-30T14:38:17.373Z", "content": "

yup if the possible cheaters are allowed to continue on kaggle, it will inspire generations of cheaters to come. Hope kaggle takes really serious measures to deal with this

", "rawMarkdown": "yup if the possible cheaters are allowed to continue on kaggle, it will inspire generations of cheaters to come. Hope kaggle takes really serious measures to deal with this", "votes": 2 }, { "id": 2122325, "postDate": "2023-01-30T21:21:15.333Z", "content": "

Guess how poteman got his solo gold? in Feedback Prize competition, he is on 8th place(with lowest amount of submissions in gold zone), while runningz is in 5th place (with fullstacked team).
\nGuess how runningz got his solo gold? in MoA competition, he is on 6th place (with lowest amount of submissions in gold zone), while poteman is in 1st(!) place (with fullstacked team)

", "rawMarkdown": "Guess how poteman got his solo gold? in Feedback Prize competition, he is on 8th place(with lowest amount of submissions in gold zone), while runningz is in 5th place (with fullstacked team).\nGuess how runningz got his solo gold? in MoA competition, he is on 6th place (with lowest amount of submissions in gold zone), while poteman is in 1st(!) place (with fullstacked team)", "votes": 11, "replies": [ { "id": 2122331, "postDate": "2023-01-30T21:27:24.640Z", "content": "

This is just a coincidence :)

\n

\"\"

\n

\"\"

", "rawMarkdown": "This is just a coincidence :)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1985486%2Fd4cb53e315e16692539d387f83bf61d9%2F2023-01-26%20152225.png?generation=1675113956551180&alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1985486%2F7614583ab897696fde2782f88d1c2627%2F2023-01-26%20152257.png?generation=1675113968115277&alt=media)", "votes": 10 } ] } ] }, { "id": 2116671, "postDate": "2023-01-26T16:29:24.700Z", "content": "

I think this is connected to @hydantess code being stolen in previous competition. Does anyone from top 20 use the same cloud platform?
\nhttps://www.kaggle.com/competitions/us-patent-phrase-to-phrase-matching/discussion/337853

", "rawMarkdown": "I think this is connected to @hydantess code being stolen in previous competition. Does anyone from top 20 use the same cloud platform?\nhttps://www.kaggle.com/competitions/us-patent-phrase-to-phrase-matching/discussion/337853", "votes": 2, "replies": [ { "id": 2116731, "postDate": "2023-01-26T17:15:34.580Z", "content": "

wtf… 👀👀

", "rawMarkdown": "wtf... 👀👀" }, { "id": 2116736, "postDate": "2023-01-26T17:19:53.847Z", "content": "

Something similar happened with the Amex competition.
\nhttps://www.kaggle.com/competitions/amex-default-prediction/discussion/348789

", "rawMarkdown": "Something similar happened with the Amex competition.\nhttps://www.kaggle.com/competitions/amex-default-prediction/discussion/348789", "replies": [ { "id": 2118381, "postDate": "2023-01-28T01:23:05.493Z", "content": "

There are some differences. I was solo then (in the money zone), and someone sitting next to me in the company stole my submission. It takes time to get out of disappointment.

", "rawMarkdown": "There are some differences. I was solo then (in the money zone), and someone sitting next to me in the company stole my submission. It takes time to get out of disappointment.", "votes": 6 }, { "id": 2118399, "postDate": "2023-01-28T01:41:45.043Z", "content": "

I'm sorry to hear that…
\nI don't want to think about cheetahs being close.

", "rawMarkdown": "I'm sorry to hear that...\nI don't want to think about cheetahs being close." } ] }, { "id": 2117091, "postDate": "2023-01-27T01:42:05.543Z", "content": "

No. It's not the same cheating group. They don't sell medals but for fame.

", "rawMarkdown": "No. It's not the same cheating group. They don't sell medals but for fame.", "votes": 4, "replies": [ { "id": 2117094, "postDate": "2023-01-27T01:47:09.507Z", "content": "

You are right. I think they could have gotten the gold medal on merit, but actually took this unglamorous approach.

", "rawMarkdown": "You are right. I think they could have gotten the gold medal on merit, but actually took this unglamorous approach.", "votes": 3 } ] } ] }, { "id": 2123479, "postDate": "2023-01-31T14:41:56.023Z", "content": "

I've thought the cheaters are filtered after the competition's end…

", "rawMarkdown": "I've thought the cheaters are filtered after the competition's end..." }, { "id": 2122338, "postDate": "2023-01-30T21:32:43.170Z", "content": "

They ruin it for everyone I guess

", "rawMarkdown": "They ruin it for everyone I guess" }, { "id": 2116611, "postDate": "2023-01-26T15:48:32.603Z", "content": "

After the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603

\n

Looking forward to reproducing the submission

", "rawMarkdown": "`After the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603`\n\nLooking forward to reproducing the submission" }, { "id": 2122235, "postDate": "2023-01-30T19:24:29.283Z", "rawMarkdown": "", "votes": -3, "isDeleted": true }, { "id": 2118748, "postDate": "2023-01-28T08:39:40.983Z", "rawMarkdown": "", "votes": 3, "isDeleted": true }, { "id": 2117167, "postDate": "2023-01-27T04:34:31.563Z", "rawMarkdown": "", "votes": 7, "isDeleted": true, "replies": [ { "id": 2117418, "postDate": "2023-01-27T09:06:31.070Z", "content": "

Why do you think that he is connected? Could you please tell us more?

", "rawMarkdown": "Why do you think that he is connected? Could you please tell us more?", "replies": [ { "id": 2117499, "postDate": "2023-01-27T10:17:08.133Z", "rawMarkdown": "", "votes": 3, "isDeleted": true }, { "id": 2117523, "postDate": "2023-01-27T10:33:27.673Z", "content": "

How do you know that he wanted to steal your solution?

", "rawMarkdown": "How do you know that he wanted to steal your solution?" }, { "id": 2117550, "postDate": "2023-01-27T10:53:05.957Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 2117567, "postDate": "2023-01-27T11:11:09.093Z", "content": "

It's really suspicious, but from my point of view this is not enough to publicly declare someone as a cheater.

", "rawMarkdown": "It's really suspicious, but from my point of view this is not enough to publicly declare someone as a cheater." } ] }, { "id": 2117428, "postDate": "2023-01-27T09:10:29.473Z", "content": "

Cheaters should be found and banned, but let's try not to point to everyone around without providing any evidences. That's basically how a lot of internet investigations were failed after a very promising start.

", "rawMarkdown": "Cheaters should be found and banned, but let's try not to point to everyone around without providing any evidences. That's basically how a lot of internet investigations were failed after a very promising start.", "votes": 1 } ] } ], "comments": [ { "id": 2119821, "author_name": "toshi_k", "author_url": "", "post_date": "2023-01-29T06:08:22.837000", "content": "

@wcukierski I hope kaggle team deal with cheaters strictly, for the fairness and the honor of all faithful kagglers.

", "votes": 15, "replies": [] }, { "id": 2117459, "author_name": "CroDoc", "author_url": "", "post_date": "2023-01-27T09:40:18.610000", "content": "

People talk more and more about suspicious activity on Kaggle. A lot of people put hard and long hours into getting a high LB position. Often it feels like an extra job (especially when starting out).

\n

Kaggle has a very high reputation in the data science/ML world. I just hope Kaggle takes good steps to assure cheating in any way is forbidden, sanctioned and made hard. There is no perfect system, but reading things like this just decreases motivation for everyone who works hard.

", "votes": 15, "replies": [] }, { "id": 2117030, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-26T23:01:35.540000", "content": "

My biggest question is do they really exist on this earth?

", "votes": 11, "replies": [ { "id": 2120919, "author_name": "Qucy Wei", "author_url": "", "post_date": "2023-01-29T22:50:28.523000", "content": "

Haha, you made me laugh

", "votes": 0, "replies": [] } ] }, { "id": 2116323, "author_name": "Makotu", "author_url": "", "post_date": "2023-01-26T12:17:09.290000", "content": "

@ilyalos

\n

Thanks for the info! I would very much like to cooperate with you.
\nWe have a suggestion as well.
\nWe have 3 different results posted by Poteman in our team's sub. Each has a score of 0.595 / 0.596 / 0.6.
\nThese results may also be an ensemble of results from other teams. (It would be a quicker story if he had used the raw other team's submissions…)
\nAfter the competition is over, let's cross-check the submissions made by the alleged members with each other!

\n

PS: Poteman's side did not give any codes for the ensemble.

", "votes": 11, "replies": [ { "id": 2116325, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-26T12:22:44.617000", "content": "

Hi @mhyodo ,
\nThank you for so quick response. Yes, probably they might be related to one of submissions that was made in our team by @bruceqdu (scores are very similar, except 0.6). Let's try to investigate it after the end of the competition.

", "votes": 4, "replies": [] } ] }, { "id": 2116331, "author_name": "Theo Viel", "author_url": "", "post_date": "2023-01-26T12:27:24.057000", "content": "

My guess would be that finding the right coefficients is not guaranteed

\n\n

However, I think that even without such info there is more than enough proof that they are cheaters.
\nMy fear is that there are more than 2 people collaborating, and that the others get away with it.

", "votes": 5, "replies": [ { "id": 2116335, "author_name": "Psi", "author_url": "", "post_date": "2023-01-26T12:29:58.500000", "content": "

Would be interesting to study the following thing:

\n\n

Actually this could be an interesting analysis across many Kagglers.

", "votes": 22, "replies": [ { "id": 2116355, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-01-26T12:43:31.090000", "content": "

It's about time to launch Kaggle fraud detection competition

", "votes": 19, "replies": [] }, { "id": 2120255, "author_name": "Jacobvvvv", "author_url": "", "post_date": "2023-01-29T13:09:22.213000", "content": "

This could be an interesting analysis

", "votes": 1, "replies": [] } ] }, { "id": 2116348, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-26T12:39:51.807000", "content": "

I think the fact that he tried to clear the message history speaks a lot.

\n

His code is very simple and I'm almost sure that he used it to get 603 submission. Again, I do not want to share it until the end of the competition, but you will see.
\nAlso at one of the screenshots he shared a list of submission file names that he tried to stack. So it simplifies the things a bit. But you're right, if more than 2 persons were involved then proposed plan will not succeed.

", "votes": 4, "replies": [ { "id": 2116354, "author_name": "", "author_url": "", "post_date": "2023-01-26T12:43:13.547000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 2116375, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T13:00:56.467000", "content": "

yes if they can't reproduce their submissions for the initial score (which they definitely can't) a 3rd guy is surely involved, helping them with the submissions. And letting them go away, means they will continue this in the future competitions too

", "votes": 0, "replies": [ { "id": 2116439, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-26T13:49:22.420000", "content": "

I think it's a bit simpler. They can reproduce it, but they just do not want to do that, because in that case they must provide us a list of submissions that they used for stacking.

", "votes": 1, "replies": [] }, { "id": 2116445, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T13:52:32.720000", "content": "

sorry for the confusion, I meant the code which they used to make the initial submissions. They can't reproduce the code they used to make subs before teaming up

", "votes": 0, "replies": [] } ] }, { "id": 2117149, "author_name": "", "author_url": "", "post_date": "2023-01-27T04:14:49.997000", "content": "", "votes": 0, "replies": [] }, { "id": 2121683, "author_name": "Gleb", "author_url": "", "post_date": "2023-01-30T13:41:49.487000", "content": "

As far as i understand none of those two shared their code, and it goes waay into other competitions. This opens up a possibility of having more people collaborating. Ill now tag some guys (that I personally saw on kaggle contributing) from other competitions, can you guys share, if poteman or runningz actually provided code in any competition you did together? @ren4yu @rohitsingh9990 @iafoss @anonamename

", "votes": 5, "replies": [ { "id": 2122415, "author_name": "Dracarys", "author_url": "", "post_date": "2023-01-30T23:16:21.390000", "content": "

Hi, @bakeryproducts my team achieved the first place in recent Feedback3 competition, poteman was part of our team. He requested me to merge with him earlier in this competition and since he was solo gold winner for feedback2 competition i agreed to merge. But after merging he got busy with his office work and wasn't able to contribute in that competition (this is what he told after we won the competition and there is no reason for us to suspect otherwise). Since he didn't contributed, never shared any code. This was my first time collaborating with him so, i can't comment on any other competition he participated but atleast for Feedback3 comp. he never shared any code with us.

", "votes": 9, "replies": [] }, { "id": 2122433, "author_name": "ONODERA", "author_url": "", "post_date": "2023-01-30T23:44:38.210000", "content": "

@rohitsingh9990 That's the 4th instance where he did nothing. I've heard every kagglers who teamed up with him won't team up with him anymore.
\nWe should share these information to avoid further casualties.

", "votes": 3, "replies": [] }, { "id": 2122491, "author_name": "anonamename", "author_url": "", "post_date": "2023-01-31T01:24:27.293000", "content": "

Hi, @bakeryproducts
\nI teamed up with him in G2Net2 competition. In this competition as well, he got busy with his office work. He did not contribute anything during competition period. After competition ended, he told me that he had tried stacking with Out-Of-Fold data I created. However, he did not share any code with me.

", "votes": 4, "replies": [] }, { "id": 2122964, "author_name": "Gleb", "author_url": "", "post_date": "2023-01-31T08:35:15.067000", "content": "

thank you @rohitsingh9990 (i mean for now there is at least some reason to suspect otherwise?) thank you @anonamename , huge info

", "votes": 2, "replies": [] } ] } ] }, { "id": 2116308, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T12:02:12.830000", "content": "

Again thank you coming out honestly, this is truly remarkable to see such honesty.

", "votes": 5, "replies": [ { "id": 2116338, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-26T12:32:56.317000", "content": "

I can say for all four of us (myself, @dempton, @bruceqdu, @hansu1017) that being a part of something unfair is really uncomfortably for us and we just cannot accept it. Hope most of the kagglers have the same position on that except very few individuals.

", "votes": 7, "replies": [ { "id": 2116376, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-26T13:01:48.063000", "content": "

yes I really hope it too

", "votes": 0, "replies": [] } ] } ] }, { "id": 2121760, "author_name": "CPMP", "author_url": "", "post_date": "2023-01-30T14:35:42.040000", "content": "

Wow. Maybe you should contact kaggle staff if not done yet, and discuss with them what evidence you need to provide to confound the suspected cheaters. Discussing evidence in front of them can help them escape. I think the mail is compliance@kaggle.com

\n

This is a really sad story for you and your honest team mates, but i hope some good will come out of it thanks to your actions. Thanks for that.

\n

FYI, a colleague of mine also spotted this and found it suspicious: \"almost every time when @runningz has a high place, @poteman is very close. \"

", "votes": 4, "replies": [ { "id": 2121765, "author_name": "NikhilMishra", "author_url": "", "post_date": "2023-01-30T14:38:17.373000", "content": "

yup if the possible cheaters are allowed to continue on kaggle, it will inspire generations of cheaters to come. Hope kaggle takes really serious measures to deal with this

", "votes": 2, "replies": [] }, { "id": 2122325, "author_name": "Gleb", "author_url": "", "post_date": "2023-01-30T21:21:15.333000", "content": "

Guess how poteman got his solo gold? in Feedback Prize competition, he is on 8th place(with lowest amount of submissions in gold zone), while runningz is in 5th place (with fullstacked team).
\nGuess how runningz got his solo gold? in MoA competition, he is on 6th place (with lowest amount of submissions in gold zone), while poteman is in 1st(!) place (with fullstacked team)

", "votes": 11, "replies": [ { "id": 2122331, "author_name": "zakopuro", "author_url": "", "post_date": "2023-01-30T21:27:24.640000", "content": "

This is just a coincidence :)

\n

\"\"

\n

\"\"

", "votes": 10, "replies": [] } ] } ] }, { "id": 2116671, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2023-01-26T16:29:24.700000", "content": "

I think this is connected to @hydantess code being stolen in previous competition. Does anyone from top 20 use the same cloud platform?
\nhttps://www.kaggle.com/competitions/us-patent-phrase-to-phrase-matching/discussion/337853

", "votes": 2, "replies": [ { "id": 2116731, "author_name": "empty", "author_url": "", "post_date": "2023-01-26T17:15:34.580000", "content": "

wtf… 👀👀

", "votes": 0, "replies": [] }, { "id": 2116736, "author_name": "zakopuro", "author_url": "", "post_date": "2023-01-26T17:19:53.847000", "content": "

Something similar happened with the Amex competition.
\nhttps://www.kaggle.com/competitions/amex-default-prediction/discussion/348789

", "votes": 0, "replies": [ { "id": 2118381, "author_name": "Qsj", "author_url": "", "post_date": "2023-01-28T01:23:05.493000", "content": "

There are some differences. I was solo then (in the money zone), and someone sitting next to me in the company stole my submission. It takes time to get out of disappointment.

", "votes": 6, "replies": [] }, { "id": 2118399, "author_name": "zakopuro", "author_url": "", "post_date": "2023-01-28T01:41:45.043000", "content": "

I'm sorry to hear that…
\nI don't want to think about cheetahs being close.

", "votes": 0, "replies": [] } ] }, { "id": 2117091, "author_name": "hyd", "author_url": "", "post_date": "2023-01-27T01:42:05.543000", "content": "

No. It's not the same cheating group. They don't sell medals but for fame.

", "votes": 4, "replies": [ { "id": 2117094, "author_name": "Bruce", "author_url": "", "post_date": "2023-01-27T01:47:09.507000", "content": "

You are right. I think they could have gotten the gold medal on merit, but actually took this unglamorous approach.

", "votes": 3, "replies": [] } ] } ] }, { "id": 2123479, "author_name": "JcJet", "author_url": "", "post_date": "2023-01-31T14:41:56.023000", "content": "

I've thought the cheaters are filtered after the competition's end…

", "votes": 0, "replies": [] }, { "id": 2122338, "author_name": "Abdullah", "author_url": "", "post_date": "2023-01-30T21:32:43.170000", "content": "

They ruin it for everyone I guess

", "votes": 0, "replies": [] }, { "id": 2116611, "author_name": "sirius", "author_url": "", "post_date": "2023-01-26T15:48:32.603000", "content": "

After the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603

\n

Looking forward to reproducing the submission

", "votes": 0, "replies": [] }, { "id": 2122235, "author_name": "", "author_url": "", "post_date": "2023-01-30T19:24:29.283000", "content": "", "votes": -3, "replies": [] }, { "id": 2118748, "author_name": "", "author_url": "", "post_date": "2023-01-28T08:39:40.983000", "content": "", "votes": 3, "replies": [] }, { "id": 2117167, "author_name": "", "author_url": "", "post_date": "2023-01-27T04:34:31.563000", "content": "", "votes": 7, "replies": [ { "id": 2117418, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-27T09:06:31.070000", "content": "

Why do you think that he is connected? Could you please tell us more?

", "votes": 0, "replies": [ { "id": 2117499, "author_name": "", "author_url": "", "post_date": "2023-01-27T10:17:08.133000", "content": "", "votes": 3, "replies": [] }, { "id": 2117523, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-27T10:33:27.673000", "content": "

How do you know that he wanted to steal your solution?

", "votes": 0, "replies": [] }, { "id": 2117550, "author_name": "", "author_url": "", "post_date": "2023-01-27T10:53:05.957000", "content": "", "votes": 1, "replies": [] }, { "id": 2117567, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-27T11:11:09.093000", "content": "

It's really suspicious, but from my point of view this is not enough to publicly declare someone as a cheater.

", "votes": 0, "replies": [] } ] }, { "id": 2117428, "author_name": "Ilia los", "author_url": "", "post_date": "2023-01-27T09:10:29.473000", "content": "

Cheaters should be found and banned, but let's try not to point to everyone around without providing any evidences. That's basically how a lot of internet investigations were failed after a very promising start.

", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "2116296": "You've already read about the situation that affected two teams on top of the leaderboard. Unfortunately my team was involved as well.\n\nYou can read these threads for more details:\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381318\nhttps://www.kaggle.com/competitions/otto-recommender-system/discussion/381321\n\nI would like to cover two main things:\n1) explain one more time what happened, but this time from my and @dempton side\n2) explain a way how we can reliably prove that these guys did share the code between each other during the competition\n\n**So what happened?**\nAt the last day before the merge deadline we were contacted by @runningz with a proposition to merge. His team's score was 601 at the moment so we rejected it, because strongly believe that could achieve such score very soon. Later that day he made a submission with a score 603 and after that we decided to accept the invitation. After the merge in a private slack space he told us that the best single model has score 599, but there are no submission with such score. We found it suspicion and decided to have a look on his history of participations. **This is not an evidence**, but just check that almost every time when @runningz has a high place, @poteman is very close. \n\nIt became even more suspicion when on the next day @bruceqdu told us that the best single model score is 597 and @runningz agreed by telling that mentioning 599 yesterday was a mistake. After that we decided to force @runningz to prove the way how he got that 603 result. He sent us several screenshots with his algorithm to ensemble the submissions and told that he cannot reproduce the 603 submission anymore.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1082778%2Fd5739c3d0a6310bcda9a2f4907878eb1%2FScreenshot%202023-01-26%20at%2010.59.29.png?generation=1674730469831845&alt=media)\nHave a look that images were deleted by him after that. Later we will find out why he did that.\n\nWe continued to push him, because after that it was much clearer that something isn't right. And we decided to go even more deeper. Of course on that particular moment we didn't know for sure that @runningz and @poteman are cheaters, but the bait worked perfectly.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1082778%2F04c09ab93b99c46dec8861330185fa2f%2FScreenshot%202023-01-26%20at%2011.57.37.png?generation=1674730672995028&alt=media)\n\nAfter that he started to clear slack history with details about his ensembling techniques. But we have saved all images and text messages before that, because expected something like that from him. And in general his behaviour was the answer for us. After that he hasn't sent any messages yet and he hasn't answer us anymore.\n\n**So what I would like to propose to try prove that these two guys are cheaters.**\nAfter the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603. We have his code for ensembling (I'm going to share it here after the end of the competition), we just need to get the correct list of submission files to stack with the right coefficients. And I'm pretty sure that the list will have at least one submission outside of our team. I think if we get the exact match for all 5015409 lines then it will be absolutely clear that it is not a coincided.\n\n@mhyodo I hope that you will be ready to collaborate on this after the end of the competition\n@pnormann please confirm that the way that I described above will lead us to ban these two gentlemen forever? Of course in case if we get exactly the same submission file. I hope that you take it all serious, because @poteman is currently top20 in kaggle competitions overall rating and @runningz is one gold medal away from competitions GM title. Cheating is a bad thing at all levels, but when such high ranked persons are involved then I think it's much worse.\n\n**And about our participation.**\nWe decided not to continue, because @runningz is holding team lead position, we've already spent 3 days on that staff and at the end we are just too demoralised. I hope kaggle will find a way to preserve us in this competitions in a some honest way. For example, to remove @runnigz and use 2 best submissions that we sent with @dempton before the merge as our team's final submissions.", "2119821": "@wcukierski I hope kaggle team deal with cheaters strictly, for the fairness and the honor of all faithful kagglers.", "2117459": "People talk more and more about suspicious activity on Kaggle. A lot of people put hard and long hours into getting a high LB position. Often it feels like an extra job (especially when starting out).\n\nKaggle has a very high reputation in the data science/ML world. I just hope Kaggle takes good steps to assure cheating in any way is forbidden, sanctioned and made hard. There is no perfect system, but reading things like this just decreases motivation for everyone who works hard.", "2117030": "My biggest question is `do they really exist on this earth?`", "2116323": "@ilyalos \n\nThanks for the info! I would very much like to cooperate with you.\nWe have a suggestion as well.\nWe have 3 different results posted by Poteman in our team's sub. Each has a score of 0.595 / 0.596 / 0.6.\nThese results may also be an ensemble of results from other teams. (It would be a quicker story if he had used the raw other team's submissions...)\nAfter the competition is over, let's cross-check the submissions made by the alleged members with each other!\n\nPS: Poteman's side did not give any codes for the ensemble.", "2116331": "My guess would be that finding the right coefficients is not guaranteed\n- The code you have may not be the one he used for the 0.603 sub\n- There may be other files involved than the ones poteman's team mates have access to.\n\nHowever, I think that even without such info there is more than enough proof that they are cheaters.\nMy fear is that there are more than 2 people collaborating, and that the others get away with it.", "2116308": "Again thank you coming out honestly, this is truly remarkable to see such honesty.", "2121760": "Wow. Maybe you should contact kaggle staff if not done yet, and discuss with them what evidence you need to provide to confound the suspected cheaters. Discussing evidence in front of them can help them escape. I think the mail is compliance@kaggle.com\n\nThis is a really sad story for you and your honest team mates, but i hope some good will come out of it thanks to your actions. Thanks for that.\n\nFYI, a colleague of mine also spotted this and found it suspicious: \"almost every time when @runningz has a high place, @poteman is very close. \"\n\n", "2116671": "I think this is connected to @hydantess code being stolen in previous competition. Does anyone from top 20 use the same cloud platform?\nhttps://www.kaggle.com/competitions/us-patent-phrase-to-phrase-matching/discussion/337853", "2123479": "I've thought the cheaters are filtered after the competition's end...", "2122338": "They ruin it for everyone I guess", "2116611": "`After the end of the competition with a help of @poteman teammates (@mhyodo and others) we could try to generate exactly the same submission file as @runningz sent to achieve 603`\n\nLooking forward to reproducing the submission", "2122235": "", "2118748": "", "2117167": "" } }