![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3015190%2Fc6bae6a9b353f55b68d142ae17093a9a%2FLOL.jpeg?generation=1577456195255323&alt=media\")
There is an increase of cheating lately on Kaggle and in many forms (like multiple fake accounts, private sharing solutions during competitions, use of external data not allowed or not disclosed on the related topic etc.) , as already highlighted in many topics
\n\nI think this may threaten the credibility of the platform in the long run, if no drastic measures are taken.
\n\nJust removing cheaters from the LB clearly doesn't seem to be a strong deterrent anymore (given the increase). And anyway a cheater will probably think: \"At best I'll slip through the cracks and Kaggle system won't spot me, at worst if I'm caught, I'll just be removed from one competition but I might continue cheating on other competitions\".
\n\nMore drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking, medal withdrawal , ban from participating on others competitions from temporary to permanent depending on the level and frequency of cheating.
", "messages": [ { "id": 699541, "postDate": "2019-12-20T15:07:18.400Z", "content": "There is an increase of cheating lately on Kaggle and in many forms (like multiple fake accounts, private sharing solutions during competitions, use of external data not allowed or not disclosed on the related topic etc.) , as already highlighted in many topics
\n\nI think this may threaten the credibility of the platform in the long run, if no drastic measures are taken.
\n\nJust removing cheaters from the LB clearly doesn't seem to be a strong deterrent anymore (given the increase). And anyway a cheater will probably think: \"At best I'll slip through the cracks and Kaggle system won't spot me, at worst if I'm caught, I'll just be removed from one competition but I might continue cheating on other competitions\".
\n\nMore drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking, medal withdrawal , ban from participating on others competitions from temporary to permanent depending on the level and frequency of cheating.
", "rawMarkdown": "\nThere is an increase of cheating lately on Kaggle and in many forms (like multiple fake accounts, private sharing solutions during competitions, use of external data not allowed or not disclosed on the related topic etc.) , as already highlighted in many topics \n\n\nI think this may threaten the credibility of the platform in the long run, if no drastic measures are taken. \n\nJust removing cheaters from the LB clearly doesn't seem to be a strong deterrent anymore (given the increase). And anyway a cheater will probably think: \"At best I'll slip through the cracks and Kaggle system won't spot me, at worst if I'm caught, I'll just be removed from one competition but I might continue cheating on other competitions\". \n\n\nMore drastic and deterrent actions against cheaters would be something like **malus on (global) points/ranking, medal withdrawal** , **ban from participating on others competitions** from temporary to permanent depending on the level and frequency of cheating.", "votes": 112 }, { "id": 704107, "postDate": "2019-12-27T04:01:28.673Z", "content": "After 73 years of life its been my experience that most actions intended to \"punish\" most often end up hurting the 95% of good folks while being only a minor pain to those who needed the punishment. This seems to have been the case through most of the history of time - according to many - all of us have been punished because one person ate the apple.
\n\nIn case you have not looked out your window for some time - the world is full of cheaters. Aside from waiting for the invention of Skynet - try not to throw the baby out with the bath water.
", "rawMarkdown": "After 73 years of life its been my experience that most actions intended to \"punish\" most often end up hurting the 95% of good folks while being only a minor pain to those who needed the punishment. This seems to have been the case through most of the history of time - according to many - all of us have been punished because one person ate the apple.\n\nIn case you have not looked out your window for some time - the world is full of cheaters. Aside from waiting for the invention of Skynet - try not to throw the baby out with the bath water.\n\n\n", "votes": 16, "replies": [ { "id": 706645, "postDate": "2019-12-30T16:48:59.793Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 699651, "postDate": "2019-12-20T17:51:51.477Z", "content": "Who is the downvoter? :)
", "rawMarkdown": "Who is the downvoter? :)", "votes": 14, "replies": [ { "id": 699758, "postDate": "2019-12-20T21:25:16.310Z", "content": "Let's hunt that guy! 😂
", "rawMarkdown": "Let's hunt that guy! 😂 ", "votes": 4 }, { "id": 699784, "postDate": "2019-12-20T23:31:45.517Z", "content": "It’s not me!
", "rawMarkdown": "It’s not me!", "votes": 1 }, { "id": 700048, "postDate": "2019-12-21T11:06:19.413Z", "content": "\n\nRob Mulla knows how to do that ;-)
\n\nhttps://www.kaggle.com/robikscube/identifying-potential-fake-accounts-on-kaggle
", "rawMarkdown": "@robikscube \n\nRob Mulla knows how to do that ;-)\n\nhttps://www.kaggle.com/robikscube/identifying-potential-fake-accounts-on-kaggle", "votes": 1 } ] }, { "id": 701204, "postDate": "2019-12-23T08:18:14.380Z", "content": "Simple cheating can be treated probably but the following ones may be a lot harder:\n - private sharing of engineered features or sharing between people that have decided to team up at competition start.\n - having a team of DS working for you \n - access to big clusters of CPUs/GPUs and/or ML softwares that give insights on features (most competitions require OSS).
\n\nFinally and most important of all, only winners have their solutions checked against the rules, the rest of gold, silver and bronze don't... That's an interesting bias between the last in money and his follower.
", "rawMarkdown": "Simple cheating can be treated probably but the following ones may be a lot harder:\n - private sharing of engineered features or sharing between people that have decided to team up at competition start.\n - having a team of DS working for you \n - access to big clusters of CPUs/GPUs and/or ML softwares that give insights on features (most competitions require OSS).\n\nFinally and most important of all, only winners have their solutions checked against the rules, the rest of gold, silver and bronze don't... That's an interesting bias between the last in money and his follower.", "votes": 10, "replies": [ { "id": 701381, "postDate": "2019-12-23T12:25:10.120Z", "content": "@ogrellier, I think Kaggle should adopt a rule of randomly checking the solution of a few people who are outside the prize range. They do not even have to tell us how many in the top-N solutions they check randomly. This can serve as a deterrent from engaging in some of these activities.
", "rawMarkdown": "@ogrellier, I think Kaggle should adopt a rule of randomly checking the solution of a few people who are outside the prize range. They do not even have to tell us how many in the top-N solutions they check randomly. This can serve as a deterrent from engaging in some of these activities.", "votes": 3 }, { "id": 701544, "postDate": "2019-12-23T15:40:34.673Z", "content": "@sheriytm I would check all gold at a minimum... but that takes time.
", "rawMarkdown": "@sheriytm I would check all gold at a minimum... but that takes time.", "votes": 3 }, { "id": 701548, "postDate": "2019-12-23T15:45:20.173Z", "content": "@ogrellier, I agree but they need to go beyond that which is why I propose a random check. That will make the gold zone even more nervous than the rest since it is so hard to get there.
", "rawMarkdown": "@ogrellier, I agree but they need to go beyond that which is why I propose a random check. That will make the gold zone even more nervous than the rest since it is so hard to get there." }, { "id": 701567, "postDate": "2019-12-23T16:05:08.347Z", "content": "How would solutions be verified?
\n\nIf you want the same process as for prize, then it requires substantial work from the participant. What if participant do not want or cannot spend the 2 man weeks of work to clean code, document it, make a presentation material, and make sure the code can be used to replicate the solution? Esp if you're a random pick near the bottom of LB.
\n\nI reacted to other comments in this topic for the same reason: while I fully support the intent, as cheaters are negatively impacting the fairness and interest of Kagle competitions, I disagree with the single sided and too simplistic cures to the problem that we can read here and there.
\n\nLimiting to gold, and asking gold medalist to provide evidence seems a reasonable trade off to me. Anything beyond is both a toll on Kaggle and on participants.
\n\nAnd don't forget that Kaggle does check suspicious accounts, and we can see they remove people all over the LB, not just near the top.
", "rawMarkdown": "How would solutions be verified?\n\nIf you want the same process as for prize, then it requires substantial work from the participant. What if participant do not want or cannot spend the 2 man weeks of work to clean code, document it, make a presentation material, and make sure the code can be used to replicate the solution? Esp if you're a random pick near the bottom of LB.\n\nI reacted to other comments in this topic for the same reason: while I fully support the intent, as cheaters are negatively impacting the fairness and interest of Kagle competitions, I disagree with the single sided and too simplistic cures to the problem that we can read here and there. \n\nLimiting to gold, and asking gold medalist to provide evidence seems a reasonable trade off to me. Anything beyond is both a toll on Kaggle and on participants.\n\nAnd don't forget that Kaggle does check suspicious accounts, and we can see they remove people all over the LB, not just near the top.", "votes": 11 }, { "id": 701597, "postDate": "2019-12-23T16:38:49.367Z", "content": "\n\n\nEsp if you're a random pick near the bottom of LB.
\n
@cpmpml, I do not mean the random check to include the bottom of the LB. Just a Top-N number of X teams to be selected randomly to check. That X can be based on an percentage M of the number of participants. And Kaggle do not need to tell us what the N is in the top-N.
\n\nHaving said that, I agree that it will be a toll on Kaggle but they do need to do something.
\n\nAs for a randomly selected participants not wanting to put in the work required to submit their code for review is our contribution to the work Kaggle team is willing to do. If one is selected and felt they do not want to do that work, they can forgoe their position in that comeptition. If what I propose is communicated to the community as the new rule of the competitions, we can abide by it.
", "rawMarkdown": "> Esp if you're a random pick near the bottom of LB.\n\n@cpmpml, I do not mean the random check to include the bottom of the LB. Just a Top-N number of X teams to be selected randomly to check. That X can be based on an percentage M of the number of participants. And Kaggle do not need to tell us what the N is in the top-N.\n\nHaving said that, I agree that it will be a toll on Kaggle but they do need to do something.\n\nAs for a randomly selected participants not wanting to put in the work required to submit their code for review is our contribution to the work Kaggle team is willing to do. If one is selected and felt they do not want to do that work, they can forgoe their position in that comeptition. If what I propose is communicated to the community as the new rule of the competitions, we can abide by it." } ] }, { "id": 700371, "postDate": "2019-12-21T22:17:53.370Z", "content": "Fully agree! Unfortunately private sharing is often hard to prove.
", "rawMarkdown": "Fully agree! Unfortunately private sharing is often hard to prove.", "votes": 7 }, { "id": 700041, "postDate": "2019-12-21T10:46:12.247Z", "content": "Yeah. Regarding kernels, we caught someone on big time plagiarism a while ago. He eventually made those kernels private. He now has a suspiciously high position with few gold medals. Besides the fact that he should have been punished, this high position is also still very annoying (I think the ranking points \"gained\" on those cheated kernels made private still count for the ranking.....).
\n\nWhile his rank is fading (he changed his name but rep still damaged), he got away with this way too easy.
\n\nLink to the discussion (Gamification Of Kernels Is Discouraging): https://www.kaggle.com/product-feedback/80571
", "rawMarkdown": "Yeah. Regarding kernels, we caught someone on big time plagiarism a while ago. He eventually made those kernels private. He now has a suspiciously high position with few gold medals. Besides the fact that he should have been punished, this high position is also still very annoying (I think the ranking points \"gained\" on those cheated kernels made private still count for the ranking.....).\n\nWhile his rank is fading (he changed his name but rep still damaged), he got away with this way too easy.\n\nLink to the discussion (Gamification Of Kernels Is Discouraging): https://www.kaggle.com/product-feedback/80571", "votes": 7, "replies": [ { "id": 700074, "postDate": "2019-12-21T12:19:31.083Z", "content": "@erikbruin I missed that discussion but yes, there is huge gamification on kernels.
\n\nA current kernel Grandmaster plagiarized almost all my kernel in house prices without direct fork and without giving me any credit. Then he won easily kernel gold medal by refreshing it every day.
\n\nAt first time I didn't complain. But reading later the comments on his kernel, someone asked him how did he optimize the models' parameters . He said he did it locally on his PC. \nI was stunned to read that and I said \"Ok it's enough ! Things are going too far here\". I told him clearly in the comments: \"Not only you plagiarized my kernel without crediting me but also you publicly claim to be the original author of my work\".
\n\nHe deleted the kernel once he become GM. But you can still see most of his kernels are just mixed plagiarism of medium posts, online courses and others kernels without any credit.
", "rawMarkdown": "@erikbruin I missed that discussion but yes, there is huge gamification on kernels. \n\nA current kernel Grandmaster plagiarized almost all my [kernel in house prices ](https://www.kaggle.com/serigne/stacked-regressions-top-4-on-leaderboard) without direct fork and without giving me any credit. Then he won easily kernel gold medal by refreshing it every day. \n\nAt first time I didn't complain. But reading later the comments on his kernel, someone asked him how did he optimize the models' parameters . He said he did it locally on his PC. \nI was stunned to read that and I said \"Ok it's enough ! Things are going too far here\". I told him clearly in the comments: \"Not only you plagiarized my kernel without crediting me but also you publicly claim to be the original author of my work\". \n\nHe deleted the kernel once he become GM. But you can still see most of his kernels are just mixed plagiarism of medium posts, online courses and others kernels without any credit. \n", "votes": 12 }, { "id": 700100, "postDate": "2019-12-21T13:10:53.523Z", "content": "OMG...I think I know who you mean. There is only one GM with that kind of kernels. Never trusted him. Will send you a PM when I get home to check the name...
", "rawMarkdown": "OMG...I think I know who you mean. There is only one GM with that kind of kernels. Never trusted him. Will send you a PM when I get home to check the name...", "votes": 4 }, { "id": 700207, "postDate": "2019-12-21T16:07:09.167Z", "content": "Edit: wrong people, sorry!
", "rawMarkdown": "Edit: wrong people, sorry!", "votes": 3 }, { "id": 700232, "postDate": "2019-12-21T16:58:43.750Z", "content": "@serigne, I think I know who u are talking about. I just checked and he has deleted some of his kernels I have seen before.
", "rawMarkdown": "@serigne, I think I know who u are talking about. I just checked and he has deleted some of his kernels I have seen before.", "votes": 2 }, { "id": 700261, "postDate": "2019-12-21T17:47:40.333Z", "content": "Those \"getting started\" competitions are home of plagiarism. I started Kaggle with those competitions and my Titanic kernel is very popular. I see copies of my kernel from different users every week.
", "rawMarkdown": "Those \"getting started\" competitions are home of plagiarism. I started Kaggle with those competitions and my Titanic kernel is very popular. I see copies of my kernel from different users every week.", "votes": 2 }, { "id": 700452, "postDate": "2019-12-22T03:44:22.167Z", "content": "Are you referring to a guy who can seamlessly achieve gold medal when he publish kernel? I'm quite impressed that he achieve such perfect gold rating (100 %) accuracy. Why most of them have become \"Plagiarist\" Grandmaster! :D
", "rawMarkdown": "Are you referring to a guy who can seamlessly achieve gold medal when he publish kernel? I'm quite impressed that he achieve such perfect gold rating (100 %) accuracy. Why most of them have become \"Plagiarist\" Grandmaster! :D", "votes": 1 }, { "id": 700565, "postDate": "2019-12-22T08:15:00.923Z", "content": "I think most people are on the right track, except for phi. I just checked and the guy that I have in mind (I don't know for sure yet but I would be very surprised if Serigne talks about a different guy) has never teamed up. Devai01, I would like to stress that we are talking about \"only' one GM. I am almost certain that all other GMs are playing it very fair.
\n\nOn top of that I am aware of 3 cheating Masters (including the one that you can find out about in the link). That leads to about 5% cheaters in kernels Master+. No, that is not what we want but also means that about 95% is playing it fair.....
", "rawMarkdown": "I think most people are on the right track, except for phi. I just checked and the guy that I have in mind (I don't know for sure yet but I would be very surprised if Serigne talks about a different guy) has never teamed up. Devai01, I would like to stress that we are talking about \"only' one GM. I am almost certain that all other GMs are playing it very fair.\n\nOn top of that I am aware of 3 cheating Masters (including the one that you can find out about in the link). That leads to about 5% cheaters in kernels Master+. No, that is not what we want but also means that about 95% is playing it fair.....", "votes": 1 }, { "id": 700625, "postDate": "2019-12-22T10:27:36.570Z", "content": "@erikbruin: Yes, you are correct. I mean those who commit plagiarism must have the title \"Plagiarist\" Grandmaster!
", "rawMarkdown": "@erikbruin: Yes, you are correct. I mean those who commit plagiarism must have the title \"Plagiarist\" Grandmaster!", "votes": 1 }, { "id": 704466, "postDate": "2019-12-27T14:16:38.537Z", "content": "Essential guides for plagiarism\n
Just happened across a person who appears to be very good at gaming the tier system and I think he has a group of buddies providing up votes. Not sure why it's made me so mad -
", "rawMarkdown": "Just happened across a person who appears to be very good at gaming the tier system and I think he has a group of buddies providing up votes. Not sure why it's made me so mad - ", "votes": 6, "replies": [ { "id": 710028, "postDate": "2020-01-04T08:34:46.717Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 710060, "postDate": "2020-01-04T09:19:52.447Z", "content": "> Not sure why it's made me so mad
\n\nBecause they are doing it blatantly
\n\n![\"l\"](\"https://i.ibb.co/KGTqsVt/lol.jpg\")
Likely nothing will happen. Last time when we hunted down that kernels time guy with tons of plagiarized kernels we really pushed to get him punished. Last thing I heard about that is that after several meetings nobody within Kaggle wanted to take ownership of the problem. We have also suggested a human moderator. Never happened either. It feels like understaffing of team Kaggle indeed (as mentioned by others).
", "rawMarkdown": "Likely nothing will happen. Last time when we hunted down that kernels time guy with tons of plagiarized kernels we really pushed to get him punished. Last thing I heard about that is that after several meetings nobody within Kaggle wanted to take ownership of the problem. We have also suggested a human moderator. Never happened either. It feels like understaffing of team Kaggle indeed (as mentioned by others). ", "votes": 1 }, { "id": 717756, "postDate": "2020-01-13T15:20:17.143Z", "content": "To be fair, @erikbruin , the last cluster of accounts upvoting one another in the discussion that I have reported has been removed by Kaggle. It took time as it was around the holidays, but they are all gone now.
", "rawMarkdown": "To be fair, @erikbruin , the last cluster of accounts upvoting one another in the discussion that I have reported has been removed by Kaggle. It took time as it was around the holidays, but they are all gone now." }, { "id": 717798, "postDate": "2020-01-13T16:25:36.790Z", "content": "Did a bit more looking at the individual who made me mad - found a posted kernel that was 100% copied from a Medium article. Used the report kernel button. Couple days later he was no longer on the site - so I think that button works.
", "rawMarkdown": "Did a bit more looking at the individual who made me mad - found a posted kernel that was 100% copied from a Medium article. Used the report kernel button. Couple days later he was no longer on the site - so I think that button works.", "votes": 4 } ] }, { "id": 700021, "postDate": "2019-12-21T09:42:31.183Z", "content": "I want to see removed teams exposed
", "rawMarkdown": "I want to see removed teams exposed", "votes": 4, "replies": [ { "id": 700312, "postDate": "2019-12-21T19:24:58.120Z", "content": "My team got removed once because a team mate used several accounts. Of course neither I nor my other team mate knew about it. We found out when our team was removed from gold zone in private LB after competition end. Do you think it would have been fair to \"expose\" us?
", "rawMarkdown": "My team got removed once because a team mate used several accounts. Of course neither I nor my other team mate knew about it. We found out when our team was removed from gold zone in private LB after competition end. Do you think it would have been fair to \"expose\" us?", "votes": 6 }, { "id": 700322, "postDate": "2019-12-21T19:37:36.033Z", "content": "There are cases like yours as you said which would be unfair. How about listing accounts associated with each other? This way teams can find out who broke the rules.
", "rawMarkdown": "There are cases like yours as you said which would be unfair. How about listing accounts associated with each other? This way teams can find out who broke the rules.", "votes": 2 }, { "id": 700351, "postDate": "2019-12-21T20:39:34.463Z", "content": "We easily found who broke the rule as the one who did it said it publicly and apologized for it. It was courageous of him.
", "rawMarkdown": "We easily found who broke the rule as the one who did it said it publicly and apologized for it. It was courageous of him.", "votes": 2 }, { "id": 701179, "postDate": "2019-12-23T07:31:32.990Z", "content": "I remember the comp 😞
\n\nI don't know if it'd possible to do it automatically or how much of manual intervention would be needed from the kaggle team-removing 1 guy, not the entire team, etc. if they were to add stricter ways of dealing with cheating.
\n\nI know Kaggle doesn't mention it publicly but they might already be working on improving (already good) cheater removals (?)
", "rawMarkdown": "I remember the comp 😞 \n\nI don't know if it'd possible to do it automatically or how much of manual intervention would be needed from the kaggle team-removing 1 guy, not the entire team, etc. if they were to add stricter ways of dealing with cheating. \n\nI know Kaggle doesn't mention it publicly but they might already be working on improving (already good) cheater removals (?)", "votes": 2 }, { "id": 701301, "postDate": "2019-12-23T10:38:07.523Z", "content": "@cpmpml, your experience in that competition is etched into my memory. Because it made me even more shy about teaming up with people. I was just about getting comfortable towards joining a team when that happened and I had to rethink.
\nIt is really sad that people break the rule knowingly hoping to get away with it. Kaggle definately should do more. Perharps randomly check the solution of a few people who are outside the prize range. This may help deter teams from engaging in some of these activities.
", "rawMarkdown": "@cpmpml, your experience in that competition is etched into my memory. Because it made me even more shy about teaming up with people. I was just about getting comfortable towards joining a team when that happened and I had to rethink. \n\nIt is really sad that people break the rule knowingly hoping to get away with it. Kaggle definately should do more. Perharps randomly check the solution of a few people who are outside the prize range. This may help deter teams from engaging in some of these activities." } ] }, { "id": 703108, "postDate": "2019-12-25T16:06:43.683Z", "content": "Kaggle provides such a great platform for Data Science enthusiasts and if someone cheats here, that person should be ashamed of himself/herself.
", "rawMarkdown": "Kaggle provides such a great platform for Data Science enthusiasts and if someone cheats here, that person should be ashamed of himself/herself.", "votes": 3 }, { "id": 705781, "postDate": "2019-12-29T13:16:53.430Z", "content": "@zaharch (nosound)
\n\nYou are right !
\n\nKaggle seems to be understaffed...
\n\nCan we help them ? I/we would for sure !
", "rawMarkdown": "@zaharch (nosound)\n\nYou are right !\n\nKaggle seems to be understaffed...\n\nCan we help them ? I/we would for sure !", "votes": 1 }, { "id": 705763, "postDate": "2019-12-29T12:49:26.373Z", "content": "I think the real problem is that Kaggle is under-staffed, it feeds many other issues. They need more people. I am not sure how we can help with it, may be an open letter to Google itself?
\n\nRight now they don't have enough resources, and the constant pressure on them is not going anywhere.
", "rawMarkdown": "I think the real problem is that Kaggle is under-staffed, it feeds many other issues. They need more people. I am not sure how we can help with it, may be an open letter to Google itself? \n\nRight now they don't have enough resources, and the constant pressure on them is not going anywhere.", "votes": 1 }, { "id": 703153, "postDate": "2019-12-25T17:27:15.480Z", "content": "And so what ?
\n\nAshamed with no sanction ?
\n\nThat’s not a way ro punish...
", "rawMarkdown": "And so what ? \n\nAshamed with no sanction ?\n\nThat’s not a way ro punish... \n", "votes": 1 }, { "id": 702340, "postDate": "2019-12-24T15:18:58.863Z", "content": "First of all , everybody is learning from everyone , i hope author also did in past. I agree if someone are taking help of anyone's material (code + content), that guy should give credit to genuine account . Don't cheat other's hardwork , give credit to him also .
", "rawMarkdown": "First of all , everybody is learning from everyone , i hope author also did in past. I agree if someone are taking help of anyone's material (code + content), that guy should give credit to genuine account . Don't cheat other's hardwork , give credit to him also . ", "votes": 1 }, { "id": 701006, "postDate": "2019-12-23T01:14:15.170Z", "content": "\n\n\nuse of external data not allowed or not disclosed on the related topic etc.)
\n
I believe this might be even more tricky to detect unless they are required to share a reproducible solution.
", "rawMarkdown": "> use of external data not allowed or not disclosed on the related topic etc.) \n\nI believe this might be even more tricky to detect unless they are required to share a reproducible solution. ", "votes": 1 }, { "id": 700607, "postDate": "2019-12-22T09:41:39.507Z", "content": "The problem with all this cheating that should trouble Kaggle is that as as serious competitor you feel less and less willing to put serious efforts in a competition if the discriminant is not anymore \"try all you can by doing your best\" but being in the right circles and be smart in ignoring and circumventing rules.
", "rawMarkdown": "The problem with all this cheating that should trouble Kaggle is that as as serious competitor you feel less and less willing to put serious efforts in a competition if the discriminant is not anymore \"try all you can by doing your best\" but being in the right circles and be smart in ignoring and circumventing rules. ", "votes": 1 }, { "id": 699598, "postDate": "2019-12-20T16:15:35.867Z", "content": "Agreed.
", "rawMarkdown": "Agreed.", "votes": 1 }, { "id": 719264, "postDate": "2020-01-15T10:22:18.047Z", "content": "Have the cheater to label all the video segmentation dataset! Please!
", "rawMarkdown": "Have the cheater to label all the video segmentation dataset! Please!" }, { "id": 701631, "postDate": "2019-12-23T17:27:53.507Z", "content": "Maybe they need a competition on here to find ways to identify/deter cheaters.
", "rawMarkdown": "Maybe they need a competition on here to find ways to identify/deter cheaters.", "votes": 2, "replies": [ { "id": 701675, "postDate": "2019-12-23T18:54:08.830Z", "content": "That will be overcrowded with cheaters :)
", "rawMarkdown": "That will be overcrowded with cheaters :)", "votes": 6 } ] }, { "id": 700536, "postDate": "2019-12-22T07:04:30.400Z", "content": "I'm a newbie on kaggle and even I experienced the same thing
", "rawMarkdown": "I'm a newbie on kaggle and even I experienced the same thing ", "votes": 2 }, { "id": 699574, "postDate": "2019-12-20T15:48:46.843Z", "content": "« More drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking... »
\n\nFully agree !
", "rawMarkdown": "« More drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking... »\n\nFully agree !", "votes": 2 }, { "id": 699562, "postDate": "2019-12-20T15:39:44.637Z", "content": "💯 % agreed. Without a credible threat, the instances of cheating are only going to increase over time.
", "rawMarkdown": "💯 % agreed. Without a credible threat, the instances of cheating are only going to increase over time.", "votes": 2 }, { "id": 705559, "postDate": "2019-12-29T03:48:04.560Z", "content": "I am curious as to the nationality of people caught up here, as I'm from Australia and we are finding an increasing amount of plagiariam in University students from overseas studying here. The National Broadcaster did a program on it https://www.abc.net.au/4corners/degrees-of-deception-promo/6398568 , its is scary how it seems to be related to cultural aspects of students.
", "rawMarkdown": "I am curious as to the nationality of people caught up here, as I'm from Australia and we are finding an increasing amount of plagiariam in University students from overseas studying here. The National Broadcaster did a program on it https://www.abc.net.au/4corners/degrees-of-deception-promo/6398568 , its is scary how it seems to be related to cultural aspects of students. ", "votes": 1 }, { "id": 701218, "postDate": "2019-12-23T08:39:23.490Z", "content": "but is cheating really that big of a problem on Kaggle? I am rather new here, but it seems as though most of the time the first place/top solutions are not some miracle/beautiful mind/golden feature shit but rather lots of very meticulous effort?
\n\ntake the solution of the first guy on the lb for example:\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122796
\n\nhe got to first place with barely any features (something that was discussed multiple times), validating similarly to the half and half kernel (like the most popular kernel this competition) and mainly by removing outliers (something that was also discussed in depth on the forum). It seems as though all the information needed to reach rank 1 was out there in the open, there really wasn`t any need to cheat. All the leaks got published too as it seems, as the rank 1 guy only used the ones publically available.
", "rawMarkdown": "but is cheating really that big of a problem on Kaggle? I am rather new here, but it seems as though most of the time the first place/top solutions are not some miracle/beautiful mind/golden feature shit but rather lots of very meticulous effort?\n\ntake the solution of the first guy on the lb for example:\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122796 \n\nhe got to first place with barely any features (something that was discussed multiple times), validating similarly to the half and half kernel (like the most popular kernel this competition) and mainly by removing outliers (something that was also discussed in depth on the forum). It seems as though all the information needed to reach rank 1 was out there in the open, there really wasn`t any need to cheat. All the leaks got published too as it seems, as the rank 1 guy only used the ones publically available.\n", "votes": 1 }, { "id": 703991, "postDate": "2019-12-26T22:59:46.563Z", "content": "I'm novice 3 days Kaggle and would like to know why people is cheating? Thank you
", "rawMarkdown": "I'm novice 3 days Kaggle and would like to know why people is cheating? Thank you", "votes": -2, "replies": [ { "id": 704753, "postDate": "2019-12-28T00:57:59.510Z", "content": "Only been around for two years but here are the reasons I would be tempted to take a bite of the apple.\n1. Submissions limited to 2 (or 5, etc) when the nature of the data lets you generate a lot more submissions that you would love to check. Cheating by duplicate accounts lets you do more submissions. This sin the result of Kaggle restrictions on limit of number - likely a fix to previous cheat or limited because of Kaggle hardware restrictions.\n2. Fairly new reason - only 30 hours of GPU time per week. If you don't have your own resources than a duplicate account cheat is way to increase GPU time. This sin clearly the result of Kaggle decision to reduce hardware costs and likely a detectable bump in number of duplicate accounts as a result of this change.\n3. Sharing between individuals who are not part of a team. Would guess that this is the result of penalty team encounters where number of submissions for team is same as for a individual. Would guess the team submission limit was a fix for previous crying game where individuals felt teams had a huge advantage. If you do plan to join as team than other reasons may exist. For me I would have loved to have given a potential team member a try-out. Sharing before forming a team would help - maybe just me, but I have had a high rate of failure for new team members being duds. If you have a very good score there is some potential that you will cheat by sharing the submission/kernel with friends.\n4. Size of the prize - not a lot of incentive to cheat when the first prize is $5000. 1 million dollars is a whole other story - having large resources of man and machine improves odds of success. This one is a hard edge case - when does \"extra\" resources cross the line or does a line even exist to be crossed. Deepfake will have professional teams and resources working on it - is that cheating? IMO its not - I would go so far as to suggest that Kaggle wants to see more professional teams but that's getting off topic.\n5. plagiarism - a number of comments in this discussion topic using this word. I have some issues with the words \"open source\" and \"plagiarism\" being used. After almost two years my skill set still very heavy on the cut and paste method. If you don't want your code used by others - than don't fri..ing share it is my motto. But I do admit to getting peeved when I see someone post an almost duplicate kernel and getting lots of up votes. Not sure who I most peeved with - the person doing the posting (who may be new and unknowing) or the bunch of folks who up vote. It seems that the \"cheating line\" exists here with regards to giving the original author full credit. The reason to cheat of course is to build your resume. The concept being that having a high Kaggle rating gives a good opportunity for jobs and/or peer praise. Having no friends and being too old for a new job makes this the only reason that does not tempt me.
\n\nOK - I have run out of steam - but think this covers 80% of the reasons to cheat. Did I miss any biggies.
", "rawMarkdown": "Only been around for two years but here are the reasons I would be tempted to take a bite of the apple.\n1. Submissions limited to 2 (or 5, etc) when the nature of the data lets you generate a lot more submissions that you would love to check. Cheating by duplicate accounts lets you do more submissions. This sin the result of Kaggle restrictions on limit of number - likely a fix to previous cheat or limited because of Kaggle hardware restrictions.\n2. Fairly new reason - only 30 hours of GPU time per week. If you don't have your own resources than a duplicate account cheat is way to increase GPU time. This sin clearly the result of Kaggle decision to reduce hardware costs and likely a detectable bump in number of duplicate accounts as a result of this change.\n3. Sharing between individuals who are not part of a team. Would guess that this is the result of penalty team encounters where number of submissions for team is same as for a individual. Would guess the team submission limit was a fix for previous crying game where individuals felt teams had a huge advantage. If you do plan to join as team than other reasons may exist. For me I would have loved to have given a potential team member a try-out. Sharing before forming a team would help - maybe just me, but I have had a high rate of failure for new team members being duds. If you have a very good score there is some potential that you will cheat by sharing the submission/kernel with friends.\n4. Size of the prize - not a lot of incentive to cheat when the first prize is $5000. 1 million dollars is a whole other story - having large resources of man and machine improves odds of success. This one is a hard edge case - when does \"extra\" resources cross the line or does a line even exist to be crossed. Deepfake will have professional teams and resources working on it - is that cheating? IMO its not - I would go so far as to suggest that Kaggle wants to see more professional teams but that's getting off topic.\n5. plagiarism - a number of comments in this discussion topic using this word. I have some issues with the words \"open source\" and \"plagiarism\" being used. After almost two years my skill set still very heavy on the cut and paste method. If you don't want your code used by others - than don't fri..ing share it is my motto. But I do admit to getting peeved when I see someone post an almost duplicate kernel and getting lots of up votes. Not sure who I most peeved with - the person doing the posting (who may be new and unknowing) or the bunch of folks who up vote. It seems that the \"cheating line\" exists here with regards to giving the original author full credit. The reason to cheat of course is to build your resume. The concept being that having a high Kaggle rating gives a good opportunity for jobs and/or peer praise. Having no friends and being too old for a new job makes this the only reason that does not tempt me. \n\nOK - I have run out of steam - but think this covers 80% of the reasons to cheat. Did I miss any biggies.", "votes": 4 }, { "id": 710273, "postDate": "2020-01-04T14:27:21.760Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 702370, "postDate": "2019-12-24T15:57:48.417Z", "content": "Competitions should be Kernel only and no outside data allowed.
", "rawMarkdown": "Competitions should be Kernel only and no outside data allowed.", "votes": -1, "replies": [ { "id": 702391, "postDate": "2019-12-24T16:17:45.370Z", "content": "Doesn't solve anything. External data can be copy pasted to notebooks.
", "rawMarkdown": "Doesn't solve anything. External data can be copy pasted to notebooks.", "votes": 2 }, { "id": 702905, "postDate": "2019-12-25T10:25:39.300Z", "content": "Kaggle can quite easily check that though.
", "rawMarkdown": "Kaggle can quite easily check that though." }, { "id": 703950, "postDate": "2019-12-26T21:17:09.337Z", "content": "Why would you copy paste ? Fitting a model using test ground truth would be enough to tune your models and use the parameters in a kernel.
\n\nKaggle team does not ask you how you tuned your models and well you can be lucky with tuning.
", "rawMarkdown": "Why would you copy paste ? Fitting a model using test ground truth would be enough to tune your models and use the parameters in a kernel. \n\nKaggle team does not ask you how you tuned your models and well you can be lucky with tuning.", "votes": 1 } ] }, { "id": 704580, "postDate": "2019-12-27T16:39:13.953Z", "content": "Wow didnt know such things exist
", "rawMarkdown": "Wow didnt know such things exist" }, { "id": 703944, "postDate": "2019-12-26T21:04:34.510Z", "content": "Nice
", "rawMarkdown": "Nice" }, { "id": 703526, "postDate": "2019-12-26T09:18:34.717Z", "content": "great
", "rawMarkdown": "great" }, { "id": 703500, "postDate": "2019-12-26T09:06:09.673Z", "content": "Yes
", "rawMarkdown": "Yes" }, { "id": 703081, "postDate": "2019-12-25T15:05:21.803Z", "content": "I think they can start by adding the RULES
", "rawMarkdown": "I think they can start by adding the ***RULES***" }, { "id": 702876, "postDate": "2019-12-25T09:38:04.493Z", "content": "Agreed.
", "rawMarkdown": "Agreed." }, { "id": 702540, "postDate": "2019-12-24T20:01:30.010Z", "content": "In order to counter cheating, \n- It could check whether somewhere in the leaderboard, if there are more than a person with also almost the same accuracy\n- it could further check each of their codes with each other. If also then they found to be cheated and action would be taken against them\n- There should only be at most o one warning, for the second time the cheater should be blacklisted and probably also their IP address, as they could just do it from their other ID.
", "rawMarkdown": "In order to counter cheating, \n- It could check whether somewhere in the leaderboard, if there are more than a person with also almost the same accuracy\n- it could further check each of their codes with each other. If also then they found to be cheated and action would be taken against them\n- There should only be at most o one warning, for the second time the cheater should be blacklisted and probably also their IP address, as they could just do it from their other ID." }, { "id": 702320, "postDate": "2019-12-24T14:40:42.493Z", "content": "I agree with you. Kaggle should block such people.
", "rawMarkdown": "I agree with you. Kaggle should block such people." }, { "id": 702185, "postDate": "2019-12-24T11:12:24.513Z", "content": "Agreed
", "rawMarkdown": "Agreed" }, { "id": 702177, "postDate": "2019-12-24T11:03:28.967Z", "content": "true
", "rawMarkdown": "true" }, { "id": 701530, "postDate": "2019-12-23T15:29:44.910Z", "content": "Totally agree!
", "rawMarkdown": "Totally agree!" }, { "id": 701118, "postDate": "2019-12-23T05:28:34.087Z", "content": "I totally agree this is completely unacceptable, especially for beginners. Personnaly I find it frustrating the fact that on some starter competitions there are like hundreds of teams/persons that only submit the answers of other previously shared notebooks or they use the ground truth solution and just mix it with a % of any of their model and still end up with greater score than any score you can get with only using machine learning models. I'm taking the example of the house pricing regression competition same goes for the titanic competition you have a lot of people doing this 0 rmle or 1 recall score. I think those participants should be removed so one can really evaluate how his model doing. At some point, I was kind of depressed because no matter what I was doing I was not getting better in my results. And I refused to see the published notebooks. Until one day I had enough and I really thought I was no good. Seeing that most notebooks take into account the real answers to get better scores and finding a topic that the real answers can be found in another previous competetion, I was relieved. But still I think they are to be removed because now I'm satisfied with my results but obviously they do not reflect how well I've done. And when someone will look at them, they won't probably know that others cheated.
", "rawMarkdown": "I totally agree this is completely unacceptable, especially for beginners. Personnaly I find it frustrating the fact that on some starter competitions there are like hundreds of teams/persons that only submit the answers of other previously shared notebooks or they use the ground truth solution and just mix it with a % of any of their model and still end up with greater score than any score you can get with only using machine learning models. I'm taking the example of the house pricing regression competition same goes for the titanic competition you have a lot of people doing this 0 rmle or 1 recall score. I think those participants should be removed so one can really evaluate how his model doing. At some point, I was kind of depressed because no matter what I was doing I was not getting better in my results. And I refused to see the published notebooks. Until one day I had enough and I really thought I was no good. Seeing that most notebooks take into account the real answers to get better scores and finding a topic that the real answers can be found in another previous competetion, I was relieved. But still I think they are to be removed because now I'm satisfied with my results but obviously they do not reflect how well I've done. And when someone will look at them, they won't probably know that others cheated. " }, { "id": 701109, "postDate": "2019-12-23T05:15:50.263Z", "content": "Because of them others deserving people's chances are going waste . Strict actions must be taken.
", "rawMarkdown": "Because of them others deserving people's chances are going waste . Strict actions must be taken." }, { "id": 700927, "postDate": "2019-12-22T20:39:30.497Z", "content": "Agree, even tho I haven't experienced it.
", "rawMarkdown": "Agree, even tho I haven't experienced it." }, { "id": 700368, "postDate": "2019-12-21T22:02:43.987Z", "content": "FYI, more and more in the DS community has stopped seeing Kaggle ranking and competitions as a credible platform due to cheating and issues brought up in this thread. It's a sad reality, and I hope the Kaggle admin team will do something about it to ensure the integrity of the competitions.
", "rawMarkdown": "FYI, more and more in the DS community has stopped seeing Kaggle ranking and competitions as a credible platform due to cheating and issues brought up in this thread. It's a sad reality, and I hope the Kaggle admin team will do something about it to ensure the integrity of the competitions." }, { "id": 700245, "postDate": "2019-12-21T17:23:26.297Z", "content": "@erikbruin thanks for sharing the link to that discussion. It looks like I was living under a rock for so long...looking at that this discussion and the few profiles that everyone has been pointing at, the Kernel medals system does seem to be prone to exploitation in it's current form. The competition and discussion rankings are more robust in that respect (although the issue of cheaters needs to be addressed to clean up the Competition progression system as well).
", "rawMarkdown": "@erikbruin thanks for sharing the link to that discussion. It looks like I was living under a rock for so long...looking at that this discussion and the few profiles that everyone has been pointing at, the Kernel medals system does seem to be prone to exploitation in it's current form. The competition and discussion rankings are more robust in that respect (although the issue of cheaters needs to be addressed to clean up the Competition progression system as well). " }, { "id": 702868, "postDate": "2019-12-25T09:28:16.390Z", "rawMarkdown": "", "isDeleted": true }, { "id": 702217, "postDate": "2019-12-24T12:11:16.543Z", "rawMarkdown": "", "isDeleted": true }, { "id": 701349, "postDate": "2019-12-23T11:36:28.343Z", "rawMarkdown": "", "isDeleted": true }, { "id": 699951, "postDate": "2019-12-21T06:47:14.450Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 704107, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2019-12-27T04:01:28.673000", "content": "After 73 years of life its been my experience that most actions intended to \"punish\" most often end up hurting the 95% of good folks while being only a minor pain to those who needed the punishment. This seems to have been the case through most of the history of time - according to many - all of us have been punished because one person ate the apple.
\n\nIn case you have not looked out your window for some time - the world is full of cheaters. Aside from waiting for the invention of Skynet - try not to throw the baby out with the bath water.
", "votes": 16, "replies": [ { "id": 706645, "author_name": "", "author_url": "", "post_date": "2019-12-30T16:48:59.793000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 699651, "author_name": "Psi", "author_url": "", "post_date": "2019-12-20T17:51:51.477000", "content": "Who is the downvoter? :)
", "votes": 14, "replies": [ { "id": 699758, "author_name": "Xuan Cao", "author_url": "", "post_date": "2019-12-20T21:25:16.310000", "content": "Let's hunt that guy! 😂
", "votes": 4, "replies": [] }, { "id": 699784, "author_name": "Kha Vo", "author_url": "", "post_date": "2019-12-20T23:31:45.517000", "content": "It’s not me!
", "votes": 1, "replies": [] }, { "id": 700048, "author_name": "Erik Bruin", "author_url": "", "post_date": "2019-12-21T11:06:19.413000", "content": "\n\nRob Mulla knows how to do that ;-)
\n\nhttps://www.kaggle.com/robikscube/identifying-potential-fake-accounts-on-kaggle
", "votes": 1, "replies": [] } ] }, { "id": 701204, "author_name": "olivier", "author_url": "", "post_date": "2019-12-23T08:18:14.380000", "content": "Simple cheating can be treated probably but the following ones may be a lot harder:\n - private sharing of engineered features or sharing between people that have decided to team up at competition start.\n - having a team of DS working for you \n - access to big clusters of CPUs/GPUs and/or ML softwares that give insights on features (most competitions require OSS).
\n\nFinally and most important of all, only winners have their solutions checked against the rules, the rest of gold, silver and bronze don't... That's an interesting bias between the last in money and his follower.
", "votes": 10, "replies": [ { "id": 701381, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-23T12:25:10.120000", "content": "@ogrellier, I think Kaggle should adopt a rule of randomly checking the solution of a few people who are outside the prize range. They do not even have to tell us how many in the top-N solutions they check randomly. This can serve as a deterrent from engaging in some of these activities.
", "votes": 3, "replies": [] }, { "id": 701544, "author_name": "olivier", "author_url": "", "post_date": "2019-12-23T15:40:34.673000", "content": "@sheriytm I would check all gold at a minimum... but that takes time.
", "votes": 3, "replies": [] }, { "id": 701548, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-23T15:45:20.173000", "content": "@ogrellier, I agree but they need to go beyond that which is why I propose a random check. That will make the gold zone even more nervous than the rest since it is so hard to get there.
", "votes": 0, "replies": [] }, { "id": 701567, "author_name": "CPMP", "author_url": "", "post_date": "2019-12-23T16:05:08.347000", "content": "How would solutions be verified?
\n\nIf you want the same process as for prize, then it requires substantial work from the participant. What if participant do not want or cannot spend the 2 man weeks of work to clean code, document it, make a presentation material, and make sure the code can be used to replicate the solution? Esp if you're a random pick near the bottom of LB.
\n\nI reacted to other comments in this topic for the same reason: while I fully support the intent, as cheaters are negatively impacting the fairness and interest of Kagle competitions, I disagree with the single sided and too simplistic cures to the problem that we can read here and there.
\n\nLimiting to gold, and asking gold medalist to provide evidence seems a reasonable trade off to me. Anything beyond is both a toll on Kaggle and on participants.
\n\nAnd don't forget that Kaggle does check suspicious accounts, and we can see they remove people all over the LB, not just near the top.
", "votes": 11, "replies": [] }, { "id": 701597, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-23T16:38:49.367000", "content": "\n\n\nEsp if you're a random pick near the bottom of LB.
\n
@cpmpml, I do not mean the random check to include the bottom of the LB. Just a Top-N number of X teams to be selected randomly to check. That X can be based on an percentage M of the number of participants. And Kaggle do not need to tell us what the N is in the top-N.
\n\nHaving said that, I agree that it will be a toll on Kaggle but they do need to do something.
\n\nAs for a randomly selected participants not wanting to put in the work required to submit their code for review is our contribution to the work Kaggle team is willing to do. If one is selected and felt they do not want to do that work, they can forgoe their position in that comeptition. If what I propose is communicated to the community as the new rule of the competitions, we can abide by it.
", "votes": 0, "replies": [] } ] }, { "id": 700371, "author_name": "Carlo", "author_url": "", "post_date": "2019-12-21T22:17:53.370000", "content": "Fully agree! Unfortunately private sharing is often hard to prove.
", "votes": 7, "replies": [] }, { "id": 700041, "author_name": "Erik Bruin", "author_url": "", "post_date": "2019-12-21T10:46:12.247000", "content": "Yeah. Regarding kernels, we caught someone on big time plagiarism a while ago. He eventually made those kernels private. He now has a suspiciously high position with few gold medals. Besides the fact that he should have been punished, this high position is also still very annoying (I think the ranking points \"gained\" on those cheated kernels made private still count for the ranking.....).
\n\nWhile his rank is fading (he changed his name but rep still damaged), he got away with this way too easy.
\n\nLink to the discussion (Gamification Of Kernels Is Discouraging): https://www.kaggle.com/product-feedback/80571
", "votes": 7, "replies": [ { "id": 700074, "author_name": "Serigne ", "author_url": "", "post_date": "2019-12-21T12:19:31.083000", "content": "@erikbruin I missed that discussion but yes, there is huge gamification on kernels.
\n\nA current kernel Grandmaster plagiarized almost all my kernel in house prices without direct fork and without giving me any credit. Then he won easily kernel gold medal by refreshing it every day.
\n\nAt first time I didn't complain. But reading later the comments on his kernel, someone asked him how did he optimize the models' parameters . He said he did it locally on his PC. \nI was stunned to read that and I said \"Ok it's enough ! Things are going too far here\". I told him clearly in the comments: \"Not only you plagiarized my kernel without crediting me but also you publicly claim to be the original author of my work\".
\n\nHe deleted the kernel once he become GM. But you can still see most of his kernels are just mixed plagiarism of medium posts, online courses and others kernels without any credit.
", "votes": 12, "replies": [] }, { "id": 700100, "author_name": "Erik Bruin", "author_url": "", "post_date": "2019-12-21T13:10:53.523000", "content": "OMG...I think I know who you mean. There is only one GM with that kind of kernels. Never trusted him. Will send you a PM when I get home to check the name...
", "votes": 4, "replies": [] }, { "id": 700207, "author_name": "Phi", "author_url": "", "post_date": "2019-12-21T16:07:09.167000", "content": "Edit: wrong people, sorry!
", "votes": 3, "replies": [] }, { "id": 700232, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-21T16:58:43.750000", "content": "@serigne, I think I know who u are talking about. I just checked and he has deleted some of his kernels I have seen before.
", "votes": 2, "replies": [] }, { "id": 700261, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2019-12-21T17:47:40.333000", "content": "Those \"getting started\" competitions are home of plagiarism. I started Kaggle with those competitions and my Titanic kernel is very popular. I see copies of my kernel from different users every week.
", "votes": 2, "replies": [] }, { "id": 700452, "author_name": "FGPC", "author_url": "", "post_date": "2019-12-22T03:44:22.167000", "content": "Are you referring to a guy who can seamlessly achieve gold medal when he publish kernel? I'm quite impressed that he achieve such perfect gold rating (100 %) accuracy. Why most of them have become \"Plagiarist\" Grandmaster! :D
", "votes": 1, "replies": [] }, { "id": 700565, "author_name": "Erik Bruin", "author_url": "", "post_date": "2019-12-22T08:15:00.923000", "content": "I think most people are on the right track, except for phi. I just checked and the guy that I have in mind (I don't know for sure yet but I would be very surprised if Serigne talks about a different guy) has never teamed up. Devai01, I would like to stress that we are talking about \"only' one GM. I am almost certain that all other GMs are playing it very fair.
\n\nOn top of that I am aware of 3 cheating Masters (including the one that you can find out about in the link). That leads to about 5% cheaters in kernels Master+. No, that is not what we want but also means that about 95% is playing it fair.....
", "votes": 1, "replies": [] }, { "id": 700625, "author_name": "FGPC", "author_url": "", "post_date": "2019-12-22T10:27:36.570000", "content": "@erikbruin: Yes, you are correct. I mean those who commit plagiarism must have the title \"Plagiarist\" Grandmaster!
", "votes": 1, "replies": [] }, { "id": 704466, "author_name": "FGPC", "author_url": "", "post_date": "2019-12-27T14:16:38.537000", "content": "Essential guides for plagiarism\n
Just happened across a person who appears to be very good at gaming the tier system and I think he has a group of buddies providing up votes. Not sure why it's made me so mad -
", "votes": 6, "replies": [ { "id": 710028, "author_name": "", "author_url": "", "post_date": "2020-01-04T08:34:46.717000", "content": "", "votes": 1, "replies": [] }, { "id": 710060, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2020-01-04T09:19:52.447000", "content": "> Not sure why it's made me so mad
\n\nBecause they are doing it blatantly
\n\n
Likely nothing will happen. Last time when we hunted down that kernels time guy with tons of plagiarized kernels we really pushed to get him punished. Last thing I heard about that is that after several meetings nobody within Kaggle wanted to take ownership of the problem. We have also suggested a human moderator. Never happened either. It feels like understaffing of team Kaggle indeed (as mentioned by others).
", "votes": 1, "replies": [] }, { "id": 717756, "author_name": "Luca Basanisi", "author_url": "", "post_date": "2020-01-13T15:20:17.143000", "content": "To be fair, @erikbruin , the last cluster of accounts upvoting one another in the discussion that I have reported has been removed by Kaggle. It took time as it was around the holidays, but they are all gone now.
", "votes": 0, "replies": [] }, { "id": 717798, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2020-01-13T16:25:36.790000", "content": "Did a bit more looking at the individual who made me mad - found a posted kernel that was 100% copied from a Medium article. Used the report kernel button. Couple days later he was no longer on the site - so I think that button works.
", "votes": 4, "replies": [] } ] }, { "id": 700021, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2019-12-21T09:42:31.183000", "content": "I want to see removed teams exposed
", "votes": 4, "replies": [ { "id": 700312, "author_name": "CPMP", "author_url": "", "post_date": "2019-12-21T19:24:58.120000", "content": "My team got removed once because a team mate used several accounts. Of course neither I nor my other team mate knew about it. We found out when our team was removed from gold zone in private LB after competition end. Do you think it would have been fair to \"expose\" us?
", "votes": 6, "replies": [] }, { "id": 700322, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2019-12-21T19:37:36.033000", "content": "There are cases like yours as you said which would be unfair. How about listing accounts associated with each other? This way teams can find out who broke the rules.
", "votes": 2, "replies": [] }, { "id": 700351, "author_name": "CPMP", "author_url": "", "post_date": "2019-12-21T20:39:34.463000", "content": "We easily found who broke the rule as the one who did it said it publicly and apologized for it. It was courageous of him.
", "votes": 2, "replies": [] }, { "id": 701179, "author_name": "Sanyam Bhutani", "author_url": "", "post_date": "2019-12-23T07:31:32.990000", "content": "I remember the comp 😞
\n\nI don't know if it'd possible to do it automatically or how much of manual intervention would be needed from the kaggle team-removing 1 guy, not the entire team, etc. if they were to add stricter ways of dealing with cheating.
\n\nI know Kaggle doesn't mention it publicly but they might already be working on improving (already good) cheater removals (?)
", "votes": 2, "replies": [] }, { "id": 701301, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-23T10:38:07.523000", "content": "@cpmpml, your experience in that competition is etched into my memory. Because it made me even more shy about teaming up with people. I was just about getting comfortable towards joining a team when that happened and I had to rethink.
\nIt is really sad that people break the rule knowingly hoping to get away with it. Kaggle definately should do more. Perharps randomly check the solution of a few people who are outside the prize range. This may help deter teams from engaging in some of these activities.
", "votes": 0, "replies": [] } ] }, { "id": 703108, "author_name": "Prithvi Sharma", "author_url": "", "post_date": "2019-12-25T16:06:43.683000", "content": "Kaggle provides such a great platform for Data Science enthusiasts and if someone cheats here, that person should be ashamed of himself/herself.
", "votes": 3, "replies": [] }, { "id": 705781, "author_name": "mezoganet", "author_url": "", "post_date": "2019-12-29T13:16:53.430000", "content": "@zaharch (nosound)
\n\nYou are right !
\n\nKaggle seems to be understaffed...
\n\nCan we help them ? I/we would for sure !
", "votes": 1, "replies": [] }, { "id": 705763, "author_name": "nosound", "author_url": "", "post_date": "2019-12-29T12:49:26.373000", "content": "I think the real problem is that Kaggle is under-staffed, it feeds many other issues. They need more people. I am not sure how we can help with it, may be an open letter to Google itself?
\n\nRight now they don't have enough resources, and the constant pressure on them is not going anywhere.
", "votes": 1, "replies": [] }, { "id": 703153, "author_name": "mezoganet", "author_url": "", "post_date": "2019-12-25T17:27:15.480000", "content": "And so what ?
\n\nAshamed with no sanction ?
\n\nThat’s not a way ro punish...
", "votes": 1, "replies": [] }, { "id": 702340, "author_name": "Manish Nath choudhary", "author_url": "", "post_date": "2019-12-24T15:18:58.863000", "content": "First of all , everybody is learning from everyone , i hope author also did in past. I agree if someone are taking help of anyone's material (code + content), that guy should give credit to genuine account . Don't cheat other's hardwork , give credit to him also .
", "votes": 1, "replies": [] }, { "id": 701006, "author_name": "Sanyam Bhutani", "author_url": "", "post_date": "2019-12-23T01:14:15.170000", "content": "\n\n\nuse of external data not allowed or not disclosed on the related topic etc.)
\n
I believe this might be even more tricky to detect unless they are required to share a reproducible solution.
", "votes": 1, "replies": [] }, { "id": 700607, "author_name": "Luca Massaron", "author_url": "", "post_date": "2019-12-22T09:41:39.507000", "content": "The problem with all this cheating that should trouble Kaggle is that as as serious competitor you feel less and less willing to put serious efforts in a competition if the discriminant is not anymore \"try all you can by doing your best\" but being in the right circles and be smart in ignoring and circumventing rules.
", "votes": 1, "replies": [] }, { "id": 699598, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2019-12-20T16:15:35.867000", "content": "Agreed.
", "votes": 1, "replies": [] }, { "id": 719264, "author_name": "Daniel Kurniadi", "author_url": "", "post_date": "2020-01-15T10:22:18.047000", "content": "Have the cheater to label all the video segmentation dataset! Please!
", "votes": 0, "replies": [] }, { "id": 701631, "author_name": "Chase Wells", "author_url": "", "post_date": "2019-12-23T17:27:53.507000", "content": "Maybe they need a competition on here to find ways to identify/deter cheaters.
", "votes": 2, "replies": [ { "id": 701675, "author_name": "Volodymyr", "author_url": "", "post_date": "2019-12-23T18:54:08.830000", "content": "That will be overcrowded with cheaters :)
", "votes": 6, "replies": [] } ] }, { "id": 700536, "author_name": "ravi tanwar", "author_url": "", "post_date": "2019-12-22T07:04:30.400000", "content": "I'm a newbie on kaggle and even I experienced the same thing
", "votes": 2, "replies": [] }, { "id": 699574, "author_name": "mezoganet", "author_url": "", "post_date": "2019-12-20T15:48:46.843000", "content": "« More drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking... »
\n\nFully agree !
", "votes": 2, "replies": [] }, { "id": 699562, "author_name": "sandy1112", "author_url": "", "post_date": "2019-12-20T15:39:44.637000", "content": "💯 % agreed. Without a credible threat, the instances of cheating are only going to increase over time.
", "votes": 2, "replies": [] }, { "id": 705559, "author_name": "DougJackson", "author_url": "", "post_date": "2019-12-29T03:48:04.560000", "content": "I am curious as to the nationality of people caught up here, as I'm from Australia and we are finding an increasing amount of plagiariam in University students from overseas studying here. The National Broadcaster did a program on it https://www.abc.net.au/4corners/degrees-of-deception-promo/6398568 , its is scary how it seems to be related to cultural aspects of students.
", "votes": 1, "replies": [] }, { "id": 701218, "author_name": "Proletheus", "author_url": "", "post_date": "2019-12-23T08:39:23.490000", "content": "but is cheating really that big of a problem on Kaggle? I am rather new here, but it seems as though most of the time the first place/top solutions are not some miracle/beautiful mind/golden feature shit but rather lots of very meticulous effort?
\n\ntake the solution of the first guy on the lb for example:\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122796
\n\nhe got to first place with barely any features (something that was discussed multiple times), validating similarly to the half and half kernel (like the most popular kernel this competition) and mainly by removing outliers (something that was also discussed in depth on the forum). It seems as though all the information needed to reach rank 1 was out there in the open, there really wasn`t any need to cheat. All the leaks got published too as it seems, as the rank 1 guy only used the ones publically available.
", "votes": 1, "replies": [] }, { "id": 703991, "author_name": "Jonatan Domènech ", "author_url": "", "post_date": "2019-12-26T22:59:46.563000", "content": "I'm novice 3 days Kaggle and would like to know why people is cheating? Thank you
", "votes": -2, "replies": [ { "id": 704753, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2019-12-28T00:57:59.510000", "content": "Only been around for two years but here are the reasons I would be tempted to take a bite of the apple.\n1. Submissions limited to 2 (or 5, etc) when the nature of the data lets you generate a lot more submissions that you would love to check. Cheating by duplicate accounts lets you do more submissions. This sin the result of Kaggle restrictions on limit of number - likely a fix to previous cheat or limited because of Kaggle hardware restrictions.\n2. Fairly new reason - only 30 hours of GPU time per week. If you don't have your own resources than a duplicate account cheat is way to increase GPU time. This sin clearly the result of Kaggle decision to reduce hardware costs and likely a detectable bump in number of duplicate accounts as a result of this change.\n3. Sharing between individuals who are not part of a team. Would guess that this is the result of penalty team encounters where number of submissions for team is same as for a individual. Would guess the team submission limit was a fix for previous crying game where individuals felt teams had a huge advantage. If you do plan to join as team than other reasons may exist. For me I would have loved to have given a potential team member a try-out. Sharing before forming a team would help - maybe just me, but I have had a high rate of failure for new team members being duds. If you have a very good score there is some potential that you will cheat by sharing the submission/kernel with friends.\n4. Size of the prize - not a lot of incentive to cheat when the first prize is $5000. 1 million dollars is a whole other story - having large resources of man and machine improves odds of success. This one is a hard edge case - when does \"extra\" resources cross the line or does a line even exist to be crossed. Deepfake will have professional teams and resources working on it - is that cheating? IMO its not - I would go so far as to suggest that Kaggle wants to see more professional teams but that's getting off topic.\n5. plagiarism - a number of comments in this discussion topic using this word. I have some issues with the words \"open source\" and \"plagiarism\" being used. After almost two years my skill set still very heavy on the cut and paste method. If you don't want your code used by others - than don't fri..ing share it is my motto. But I do admit to getting peeved when I see someone post an almost duplicate kernel and getting lots of up votes. Not sure who I most peeved with - the person doing the posting (who may be new and unknowing) or the bunch of folks who up vote. It seems that the \"cheating line\" exists here with regards to giving the original author full credit. The reason to cheat of course is to build your resume. The concept being that having a high Kaggle rating gives a good opportunity for jobs and/or peer praise. Having no friends and being too old for a new job makes this the only reason that does not tempt me.
\n\nOK - I have run out of steam - but think this covers 80% of the reasons to cheat. Did I miss any biggies.
", "votes": 4, "replies": [] }, { "id": 710273, "author_name": "", "author_url": "", "post_date": "2020-01-04T14:27:21.760000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 702370, "author_name": "Macit Giray Gökırmak", "author_url": "", "post_date": "2019-12-24T15:57:48.417000", "content": "Competitions should be Kernel only and no outside data allowed.
", "votes": -1, "replies": [ { "id": 702391, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2019-12-24T16:17:45.370000", "content": "Doesn't solve anything. External data can be copy pasted to notebooks.
", "votes": 2, "replies": [] }, { "id": 702905, "author_name": "Psi", "author_url": "", "post_date": "2019-12-25T10:25:39.300000", "content": "Kaggle can quite easily check that though.
", "votes": 0, "replies": [] }, { "id": 703950, "author_name": "olivier", "author_url": "", "post_date": "2019-12-26T21:17:09.337000", "content": "Why would you copy paste ? Fitting a model using test ground truth would be enough to tune your models and use the parameters in a kernel.
\n\nKaggle team does not ask you how you tuned your models and well you can be lucky with tuning.
", "votes": 1, "replies": [] } ] }, { "id": 704580, "author_name": "Alvin", "author_url": "", "post_date": "2019-12-27T16:39:13.953000", "content": "Wow didnt know such things exist
", "votes": 0, "replies": [] }, { "id": 703944, "author_name": "Caleb Lawhorne", "author_url": "", "post_date": "2019-12-26T21:04:34.510000", "content": "Nice
", "votes": 0, "replies": [] }, { "id": 703526, "author_name": "qianerba", "author_url": "", "post_date": "2019-12-26T09:18:34.717000", "content": "great
", "votes": 0, "replies": [] }, { "id": 703500, "author_name": "Sayam Rakshit", "author_url": "", "post_date": "2019-12-26T09:06:09.673000", "content": "Yes
", "votes": 0, "replies": [] }, { "id": 703081, "author_name": "HishamElamir", "author_url": "", "post_date": "2019-12-25T15:05:21.803000", "content": "I think they can start by adding the RULES
", "votes": 0, "replies": [] }, { "id": 702876, "author_name": "", "author_url": "", "post_date": "2019-12-25T09:38:04.493000", "content": "", "votes": 0, "replies": [] }, { "id": 702540, "author_name": "", "author_url": "", "post_date": "2019-12-24T20:01:30.010000", "content": "", "votes": 0, "replies": [] }, { "id": 702320, "author_name": "", "author_url": "", "post_date": "2019-12-24T14:40:42.493000", "content": "", "votes": 0, "replies": [] }, { "id": 702185, "author_name": "", "author_url": "", "post_date": "2019-12-24T11:12:24.513000", "content": "", "votes": 0, "replies": [] }, { "id": 702177, "author_name": "", "author_url": "", "post_date": "2019-12-24T11:03:28.967000", "content": "", "votes": 0, "replies": [] }, { "id": 701530, "author_name": "", "author_url": "", "post_date": "2019-12-23T15:29:44.910000", "content": "", "votes": 0, "replies": [] }, { "id": 701118, "author_name": "", "author_url": "", "post_date": "2019-12-23T05:28:34.087000", "content": "", "votes": 0, "replies": [] }, { "id": 701109, "author_name": "", "author_url": "", "post_date": "2019-12-23T05:15:50.263000", "content": "", "votes": 0, "replies": [] }, { "id": 700927, "author_name": "", "author_url": "", "post_date": "2019-12-22T20:39:30.497000", "content": "", "votes": 0, "replies": [] }, { "id": 700368, "author_name": "", "author_url": "", "post_date": "2019-12-21T22:02:43.987000", "content": "", "votes": 0, "replies": [] }, { "id": 700245, "author_name": "", "author_url": "", "post_date": "2019-12-21T17:23:26.297000", "content": "", "votes": 0, "replies": [] }, { "id": 702868, "author_name": "", "author_url": "", "post_date": "2019-12-25T09:28:16.390000", "content": "", "votes": 0, "replies": [] }, { "id": 702217, "author_name": "", "author_url": "", "post_date": "2019-12-24T12:11:16.543000", "content": "", "votes": 0, "replies": [] }, { "id": 701349, "author_name": "", "author_url": "", "post_date": "2019-12-23T11:36:28.343000", "content": "", "votes": 0, "replies": [] }, { "id": 699951, "author_name": "", "author_url": "", "post_date": "2019-12-21T06:47:14.450000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "699541": "\nThere is an increase of cheating lately on Kaggle and in many forms (like multiple fake accounts, private sharing solutions during competitions, use of external data not allowed or not disclosed on the related topic etc.) , as already highlighted in many topics \n\n\nI think this may threaten the credibility of the platform in the long run, if no drastic measures are taken. \n\nJust removing cheaters from the LB clearly doesn't seem to be a strong deterrent anymore (given the increase). And anyway a cheater will probably think: \"At best I'll slip through the cracks and Kaggle system won't spot me, at worst if I'm caught, I'll just be removed from one competition but I might continue cheating on other competitions\". \n\n\nMore drastic and deterrent actions against cheaters would be something like **malus on (global) points/ranking, medal withdrawal** , **ban from participating on others competitions** from temporary to permanent depending on the level and frequency of cheating.", "704107": "After 73 years of life its been my experience that most actions intended to \"punish\" most often end up hurting the 95% of good folks while being only a minor pain to those who needed the punishment. This seems to have been the case through most of the history of time - according to many - all of us have been punished because one person ate the apple.\n\nIn case you have not looked out your window for some time - the world is full of cheaters. Aside from waiting for the invention of Skynet - try not to throw the baby out with the bath water.\n\n\n", "699651": "Who is the downvoter? :)", "701204": "Simple cheating can be treated probably but the following ones may be a lot harder:\n - private sharing of engineered features or sharing between people that have decided to team up at competition start.\n - having a team of DS working for you \n - access to big clusters of CPUs/GPUs and/or ML softwares that give insights on features (most competitions require OSS).\n\nFinally and most important of all, only winners have their solutions checked against the rules, the rest of gold, silver and bronze don't... That's an interesting bias between the last in money and his follower.", "700371": "Fully agree! Unfortunately private sharing is often hard to prove.", "700041": "Yeah. Regarding kernels, we caught someone on big time plagiarism a while ago. He eventually made those kernels private. He now has a suspiciously high position with few gold medals. Besides the fact that he should have been punished, this high position is also still very annoying (I think the ranking points \"gained\" on those cheated kernels made private still count for the ranking.....).\n\nWhile his rank is fading (he changed his name but rep still damaged), he got away with this way too easy.\n\nLink to the discussion (Gamification Of Kernels Is Discouraging): https://www.kaggle.com/product-feedback/80571", "709829": "Just happened across a person who appears to be very good at gaming the tier system and I think he has a group of buddies providing up votes. Not sure why it's made me so mad - ", "700021": "I want to see removed teams exposed", "703108": "Kaggle provides such a great platform for Data Science enthusiasts and if someone cheats here, that person should be ashamed of himself/herself.", "705781": "@zaharch (nosound)\n\nYou are right !\n\nKaggle seems to be understaffed...\n\nCan we help them ? I/we would for sure !", "705763": "I think the real problem is that Kaggle is under-staffed, it feeds many other issues. They need more people. I am not sure how we can help with it, may be an open letter to Google itself? \n\nRight now they don't have enough resources, and the constant pressure on them is not going anywhere.", "703153": "And so what ? \n\nAshamed with no sanction ?\n\nThat’s not a way ro punish... \n", "702340": "First of all , everybody is learning from everyone , i hope author also did in past. I agree if someone are taking help of anyone's material (code + content), that guy should give credit to genuine account . Don't cheat other's hardwork , give credit to him also . ", "701006": "> use of external data not allowed or not disclosed on the related topic etc.) \n\nI believe this might be even more tricky to detect unless they are required to share a reproducible solution. ", "700607": "The problem with all this cheating that should trouble Kaggle is that as as serious competitor you feel less and less willing to put serious efforts in a competition if the discriminant is not anymore \"try all you can by doing your best\" but being in the right circles and be smart in ignoring and circumventing rules. ", "699598": "Agreed.", "719264": "Have the cheater to label all the video segmentation dataset! Please!", "701631": "Maybe they need a competition on here to find ways to identify/deter cheaters.", "700536": "I'm a newbie on kaggle and even I experienced the same thing ", "699574": "« More drastic and deterrent actions against cheaters would be something like malus on (global) points/ranking... »\n\nFully agree !", "699562": "💯 % agreed. Without a credible threat, the instances of cheating are only going to increase over time.", "705559": "I am curious as to the nationality of people caught up here, as I'm from Australia and we are finding an increasing amount of plagiariam in University students from overseas studying here. The National Broadcaster did a program on it https://www.abc.net.au/4corners/degrees-of-deception-promo/6398568 , its is scary how it seems to be related to cultural aspects of students. ", "701218": "but is cheating really that big of a problem on Kaggle? I am rather new here, but it seems as though most of the time the first place/top solutions are not some miracle/beautiful mind/golden feature shit but rather lots of very meticulous effort?\n\ntake the solution of the first guy on the lb for example:\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122796 \n\nhe got to first place with barely any features (something that was discussed multiple times), validating similarly to the half and half kernel (like the most popular kernel this competition) and mainly by removing outliers (something that was also discussed in depth on the forum). It seems as though all the information needed to reach rank 1 was out there in the open, there really wasn`t any need to cheat. All the leaks got published too as it seems, as the rank 1 guy only used the ones publically available.\n", "703991": "I'm novice 3 days Kaggle and would like to know why people is cheating? Thank you", "702370": "Competitions should be Kernel only and no outside data allowed.", "704580": "Wow didnt know such things exist", "703944": "Nice", "703526": "great", "703500": "Yes", "703081": "I think they can start by adding the ***RULES***", "702876": "Agreed.", "702540": "In order to counter cheating, \n- It could check whether somewhere in the leaderboard, if there are more than a person with also almost the same accuracy\n- it could further check each of their codes with each other. If also then they found to be cheated and action would be taken against them\n- There should only be at most o one warning, for the second time the cheater should be blacklisted and probably also their IP address, as they could just do it from their other ID.", "702320": "I agree with you. Kaggle should block such people.", "702185": "Agreed", "702177": "true", "701530": "Totally agree!", "701118": "I totally agree this is completely unacceptable, especially for beginners. Personnaly I find it frustrating the fact that on some starter competitions there are like hundreds of teams/persons that only submit the answers of other previously shared notebooks or they use the ground truth solution and just mix it with a % of any of their model and still end up with greater score than any score you can get with only using machine learning models. I'm taking the example of the house pricing regression competition same goes for the titanic competition you have a lot of people doing this 0 rmle or 1 recall score. I think those participants should be removed so one can really evaluate how his model doing. At some point, I was kind of depressed because no matter what I was doing I was not getting better in my results. And I refused to see the published notebooks. Until one day I had enough and I really thought I was no good. Seeing that most notebooks take into account the real answers to get better scores and finding a topic that the real answers can be found in another previous competetion, I was relieved. But still I think they are to be removed because now I'm satisfied with my results but obviously they do not reflect how well I've done. And when someone will look at them, they won't probably know that others cheated. ", "701109": "Because of them others deserving people's chances are going waste . Strict actions must be taken.", "700927": "Agree, even tho I haven't experienced it.", "700368": "FYI, more and more in the DS community has stopped seeing Kaggle ranking and competitions as a credible platform due to cheating and issues brought up in this thread. It's a sad reality, and I hope the Kaggle admin team will do something about it to ensure the integrity of the competitions.", "700245": "@erikbruin thanks for sharing the link to that discussion. It looks like I was living under a rock for so long...looking at that this discussion and the few profiles that everyone has been pointing at, the Kernel medals system does seem to be prone to exploitation in it's current form. The competition and discussion rankings are more robust in that respect (although the issue of cheaters needs to be addressed to clean up the Competition progression system as well). ", "702868": "", "702217": "", "701349": "", "699951": "" } }