And we are only 11 places away from the Bronze. I think something needs to be changed!
\n\nEdit: 6 places away from the Bronze and looks like this is final.
\n\nEdit: Check this out, the leaderboard progression animation: https://www.kaggle.com/inversion/talkingdata-leaderboard-progression/code
", "messages": [ { "id": "324856", "postDate": "05/08/2018 00:44:09", "content": "And we are only 11 places away from the Bronze. I think something needs to be changed!
\n\nEdit: 6 places away from the Bronze and looks like this is final.
\n\nEdit: Check this out, the leaderboard progression animation: https://www.kaggle.com/inversion/talkingdata-leaderboard-progression/code
", "rawMarkdown": "And we are only 11 places away from the Bronze. I think something needs to be changed!\n\nEdit: 6 places away from the Bronze and looks like this is final.\n\nEdit: Check this out, the leaderboard progression animation: https://www.kaggle.com/inversion/talkingdata-leaderboard-progression/code", "votes": null }, { "id": "324883", "postDate": "05/08/2018 01:11:16", "content": "This is a shame!
", "rawMarkdown": "This is a shame!", "votes": null }, { "id": "324893", "postDate": "05/08/2018 01:22:42", "content": "And this is just the copies. Imagine how many people blended their results with those last minute kernels. Lesson learnt, if I participate in a Kaggle competition again, I need to wait for the last minute kernels and blend my work with them.
", "rawMarkdown": "And this is just the copies. Imagine how many people blended their results with those last minute kernels. Lesson learnt, if I participate in a Kaggle competition again, I need to wait for the last minute kernels and blend my work with them.", "votes": null }, { "id": "324905", "postDate": "05/08/2018 01:27:40", "content": "It's ironic that a Fraud Detection Challenge had such rampant \"fraud\" in the submissions.
", "rawMarkdown": "It's ironic that a Fraud Detection Challenge had such rampant \"fraud\" in the submissions.", "votes": null }, { "id": "324906", "postDate": "05/08/2018 01:27:53", "content": "Exactly, that's totally meaningless.
", "rawMarkdown": "Exactly, that's totally meaningless.", "votes": null }, { "id": "324965", "postDate": "05/08/2018 02:19:34", "content": "It's so sad to see someone work hard to get 09800,and only a sharing of 09811 kernel ruins all
", "rawMarkdown": "It's so sad to see someone work hard to get 09800,and only a sharing of 09811 kernel ruins all", "votes": null }, { "id": "324970", "postDate": "05/08/2018 02:26:17", "content": "Next time maybe we should just wait for the last day, dowload some submissions, do some blendings, drink a coffee, win medals. Why would we pay months of time/effort to achieve nothing? Ironic.
", "rawMarkdown": "Next time maybe we should just wait for the last day, dowload some submissions, do some blendings, drink a coffee, win medals. Why would we pay months of time/effort to achieve nothing? Ironic.", "votes": null }, { "id": "324974", "postDate": "05/08/2018 02:29:10", "content": "all you need is to stay night for the last day, and wait for last minute to blend some kernels.
", "rawMarkdown": "all you need is to stay night for the last day, and wait for last minute to blend some kernels.", "votes": null }, { "id": "324979", "postDate": "05/08/2018 02:31:31", "content": "Just wait to last day, 'do some blendings, drink a coffee, win medals.' from @Wenjie Bai
", "rawMarkdown": "Just wait to last day, 'do some blendings, drink a coffee, win medals.' from @Wenjie Bai", "votes": null }, { "id": "324995", "postDate": "05/08/2018 02:57:31", "content": "I'm going to repeat a suggestion I made here: submissions from public kernels should be disabled during the last week. (You could still fork public kernels and submit them privately, but people wouldn't be able to create public kernels with verified scores, so this kind of rule would tend to short-circuit the last minute spoiler process.)
", "rawMarkdown": "I'm going to repeat a suggestion I made [here][1]: submissions from public kernels should be disabled during the last week. (You could still fork public kernels and submit them privately, but people wouldn't be able to create public kernels with verified scores, so this kind of rule would tend to short-circuit the last minute spoiler process.) \n\n [1]: https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/54400#315120", "votes": null }, { "id": "325005", "postDate": "05/08/2018 03:09:59", "content": "I am pasting the same what I had suggested earlier here
\nMy suggestion to @Kaggle is
\nto make an intermediate dead line like ( one/two month after launch of competition) ,
Till this date kernels can be made public.
\nlater this date all new kernels will become private kernels.
\nThese can be made public post competition.
\nBy doing this Blending can still be be done using past public kernels also it encourages Diverse solutions nearing to competition closure. Thank you
That is an excellent suggestion and I vote for it! It will be the best solution.
", "rawMarkdown": "That is an excellent suggestion and I vote for it! It will be the best solution.", "votes": null }, { "id": "325024", "postDate": "05/08/2018 03:29:40", "content": "Hope to see it happen in my second competition. If it doesn't happen, I will be more prepared.\nA huge thing learned is that I need to spend more time reading discussions and kernels.
", "rawMarkdown": "Hope to see it happen in my second competition. If it doesn't happen, I will be more prepared.\nA huge thing learned is that I need to spend more time reading discussions and kernels.", "votes": null }, { "id": "325029", "postDate": "05/08/2018 03:34:04", "content": "I was not comfortable at the beginning, but after final results announced, Top competitor ranks did not change significantly.\nAnd then I realized, if my model were good enough, It wouldn't affect me at all, so to me, there are still many things to learn.
", "rawMarkdown": "I was not comfortable at the beginning, but after final results announced, Top competitor ranks did not change significantly.\nAnd then I realized, if my model were good enough, It wouldn't affect me at all, so to me, there are still many things to learn.", "votes": null }, { "id": "325035", "postDate": "05/08/2018 03:43:59", "content": "Totally agree, I know that after some point those public kernels wouldn't heart much. But if submitting exactly the same file as provided by last-minute public kernel results in Silver, there is surely something wrong with it.
\n\nI had 3 models scoring 0.9814 and averaged them, submitted at 5:03 pm. If I was on time, that score would provide us with the 222nd place on the private leaderboard and last minute kernels wouldn't heart us much. Even then, we would have a chance to win silver instead of bronze if not the last-minute kernels. But I am more frustrated not because of my unpunctuality but because of the unfair last minute sharings.
\n\nFrom not being punctual, I learned to be more punctual. From last-minute public sharings, I learned that cheating can be legalized. I don't think that the second learning adds any value to my data science skills.
", "rawMarkdown": "Totally agree, I know that after some point those public kernels wouldn't heart much. But if submitting exactly the same file as provided by last-minute public kernel results in Silver, there is surely something wrong with it.\n\nI had 3 models scoring 0.9814 and averaged them, submitted at 5:03 pm. If I was on time, that score would provide us with the 222nd place on the private leaderboard and last minute kernels wouldn't heart us much. Even then, we would have a chance to win silver instead of bronze if not the last-minute kernels. But I am more frustrated not because of my unpunctuality but because of the unfair last minute sharings. \n\nFrom not being punctual, I learned to be more punctual. From last-minute public sharings, I learned that cheating can be legalized. I don't think that the second learning adds any value to my data science skills.", "votes": null }, { "id": "325083", "postDate": "05/08/2018 05:24:55", "content": "Very disheartening. Some learnings but feels like a big waste of effort and money :(
", "rawMarkdown": "Very disheartening. Some learnings but feels like a big waste of effort and money :(", "votes": null }, { "id": "325093", "postDate": "05/08/2018 05:44:01", "content": "Totally agree with you, if they use true ability to win me i shall be gald to receive learnings, but what kind of learnings this one is counted for? 5hours per day for 40days, i feel now they are completely waste.
", "rawMarkdown": "Totally agree with you, if they use true ability to win me i shall be gald to receive learnings, but what kind of learnings this one is counted for? 5hours per day for 40days, i feel now they are completely waste.", "votes": null }, { "id": "325126", "postDate": "05/08/2018 06:36:14", "content": "That's part of the solution, but not entirely - if I remember correctly, the spoiler in Instacart Market Basket competition was shared via github. And the impact was huge, I had been in the top silver range but at the end could barely squeeze out a bronze.
", "rawMarkdown": "That's part of the solution, but not entirely - if I remember correctly, the spoiler in Instacart Market Basket competition was shared via github. And the impact was huge, I had been in the top silver range but at the end could barely squeeze out a bronze.", "votes": null }, { "id": "325153", "postDate": "05/08/2018 07:15:48", "content": "I thought till the end , that all this public kernels were overfitted and believed in my solution. It was a mistake(
", "rawMarkdown": "I thought till the end , that all this public kernels were overfitted and believed in my solution. It was a mistake(", "votes": null }, { "id": "325188", "postDate": "05/08/2018 07:53:12", "content": "It's not the first time, and it won't be the last time. Unfortunately.
\n\nThough, it is not hard to scan through the duplicated results and removed them except the first submission. You can't have 99.9% similarity if it is not plagiarism.
", "rawMarkdown": "It's not the first time, and it won't be the last time. Unfortunately.\n\nThough, it is not hard to scan through the duplicated results and removed them except the first submission. You can't have 99.9% similarity if it is not plagiarism.", "votes": null }, { "id": "325207", "postDate": "05/08/2018 08:14:40", "content": "I don't think there are any direct copied submissions in medals. I've just tested it and it returns a private score of 0.9819167 which is below yours. But, as you said, it may have been used anyways for blending.
", "rawMarkdown": "I don't think there are any direct copied submissions in medals. I've just tested it and it returns a private score of 0.9819167 which is below yours. But, as you said, it may have been used anyways for blending.", "votes": null }, { "id": "325303", "postDate": "05/08/2018 09:39:08", "content": "I completely understand the bitterness, resentment and angerness in this competition. I was not in the bronze medal until the last minute. I was about to quit and I saw this kernel. I decided to submit it. I was in a situation where I had nothing to loose. Should this kernel have overfitted the public score dataset, all the crowds would have been wiped out. It turns out the other way...
\n\nBut at the end of the day, this is just a game and the real thing is to learn.... Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...
\n\nLast advice, it seems that if you are not a top kaggler performer, it is very wise to keep your last day submission in case something happens. I did not follow you and i was tracking closely Wenjie Bai and found that she was very clever and inspiring. At the end of the day, this medal is neither a diploma or real money and we should not put too much emphasis on it!!!
\n\nFinally, the Andy Harless suggestion, no kernel the last day seems wise to me.\nAll the best to you and keep it up with kaggle
", "rawMarkdown": "I completely understand the bitterness, resentment and angerness in this competition. I was not in the bronze medal until the last minute. I was about to quit and I saw this kernel. I decided to submit it. I was in a situation where I had nothing to loose. Should this kernel have overfitted the public score dataset, all the crowds would have been wiped out. It turns out the other way...\n\nBut at the end of the day, this is just a game and the real thing is to learn.... Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...\n\nLast advice, it seems that if you are not a top kaggler performer, it is very wise to keep your last day submission in case something happens. I did not follow you and i was tracking closely Wenjie Bai and found that she was very clever and inspiring. At the end of the day, this medal is neither a diploma or real money and we should not put too much emphasis on it!!!\n\nFinally, the Andy Harless suggestion, no kernel the last day seems wise to me.\nAll the best to you and keep it up with kaggle", "votes": null }, { "id": "325463", "postDate": "05/08/2018 12:44:58", "content": "just submit there isn’t a need to blend to get silver anyways why any extra effort ;)
", "rawMarkdown": "just submit there isn’t a need to blend to get silver anyways why any extra effort ;)", "votes": null }, { "id": "325471", "postDate": "05/08/2018 12:55:24", "content": "This is by far the worst kaggle has been in my four years - take comfort it will be better next time
\n\nThe only way is up ;)
", "rawMarkdown": "This is by far the worst kaggle has been in my four years - take comfort it will be better next time\n\nThe only way is up ;)", "votes": null }, { "id": "325479", "postDate": "05/08/2018 13:03:10", "content": "You tested only one of them. Check the scores. Exact same scores have almost no chance of coming from different submission files.
", "rawMarkdown": "You tested only one of them. Check the scores. Exact same scores have almost no chance of coming from different submission files.", "votes": null }, { "id": "325486", "postDate": "05/08/2018 13:14:47", "content": "I will disagree with you. Not with the part of learning. Yes, I learned a lot during the past month. And no one can take the knowledge away from me.
\n\nBut if it wasn't for the medal, you wouldn't take a chance and submit it in the first place. Have you seen how many people participate in research or playground competitions compared to the featured competitions? Yes, a huge difference. And that's because of the medals and prizes!
\n\nKaggle's fame of being a great place for learning is because of the medals and prizes. Let's be honest, we, human beings, like to be praised for our achievements. So let's not be too modest and say that all that matters is the knowledge. First time competing in Kaggle and I had to be in the Bronze, even had a chance for Silver, and that was the motivation which was taken away from me because of the poor construction of the rules.
", "rawMarkdown": "I will disagree with you. Not with the part of learning. Yes, I learned a lot during the past month. And no one can take the knowledge away from me.\n\nBut if it wasn't for the medal, you wouldn't take a chance and submit it in the first place. Have you seen how many people participate in research or playground competitions compared to the featured competitions? Yes, a huge difference. And that's because of the medals and prizes!\n\nKaggle's fame of being a great place for learning is because of the medals and prizes. Let's be honest, we, human beings, like to be praised for our achievements. So let's not be too modest and say that all that matters is the knowledge. First time competing in Kaggle and I had to be in the Bronze, even had a chance for Silver, and that was the motivation which was taken away from me because of the poor construction of the rules.", "votes": null }, { "id": "325496", "postDate": "05/08/2018 13:25:15", "content": "@Araks, i like your view, and i'd like to regroup guys like you for next competition.
", "rawMarkdown": "Araks, i like your view, and i'd like to regroup guys like you for next competition.", "votes": null }, { "id": "325513", "postDate": "05/08/2018 13:43:22", "content": "Started following you. I have decided to try next competitions solo at least until the merger deadline. We will see how it goes. Will keep in touch with you.
", "rawMarkdown": "Started following you. I have decided to try next competitions solo at least until the merger deadline. We will see how it goes. Will keep in touch with you.", "votes": null }, { "id": "325518", "postDate": "05/08/2018 13:48:56", "content": "@Aljaž, that's awful! Sharing via GitHub is even against the rules. Weren't they disqualified?
", "rawMarkdown": "Aljaž, that's awful! Sharing via GitHub is even against the rules. Weren't they disqualified?", "votes": null }, { "id": "325524", "postDate": "05/08/2018 13:54:09", "content": "If it can be of any solace, think of me, just a single place away from the bronze medal :-( A couple of days ago I have been at position 142, actually, after literally throwing money at AWS instances in order to achieve a performing model.
\n\nStoically, I do not say anything about such a surreal competition's ending because I've seen really much far worse situations in 118 competitions I participated in. And I assure you, nothing has ever changed in spite of much protest and discussion. Such is the reality of participating to such competitions. Enjoy what you have learned, really, points and medals are just ephemeral.
", "rawMarkdown": "If it can be of any solace, think of me, just a single place away from the bronze medal :-( A couple of days ago I have been at position 142, actually, after literally throwing money at AWS instances in order to achieve a performing model.\n\nStoically, I do not say anything about such a surreal competition's ending because I've seen really much far worse situations in 118 competitions I participated in. And I assure you, nothing has ever changed in spite of much protest and discussion. Such is the reality of participating to such competitions. Enjoy what you have learned, really, points and medals are just ephemeral.", "votes": null }, { "id": "325526", "postDate": "05/08/2018 13:54:19", "content": "@Scirpus, great spirit, perfect :)
", "rawMarkdown": "Scirpus, great spirit, perfect :)", "votes": null }, { "id": "325530", "postDate": "05/08/2018 14:05:22", "content": "@Luca, thanks for being here. I am sure at least one person will be disqualified and you will get your earned medal. Yesterday we were 407th, now I see that we are 406th. So there is a great hope that you will get Bronze.
\n\nThis was a surprise for me as this is my first competition, but I spent quite some time reading other discussions yesterday and realized that it's very normal. Kind of a new motivation for me, to work harder and end in a better place so that this kind of things don't affect my results.
\n\nStill, have hopes for future improvements of the rules.
", "rawMarkdown": "Luca, thanks for being here. I am sure at least one person will be disqualified and you will get your earned medal. Yesterday we were 407th, now I see that we are 406th. So there is a great hope that you will get Bronze. \n\nThis was a surprise for me as this is my first competition, but I spent quite some time reading other discussions yesterday and realized that it's very normal. Kind of a new motivation for me, to work harder and end in a better place so that this kind of things don't affect my results. \n\nStill, have hopes for future improvements of the rules.", "votes": null }, { "id": "325577", "postDate": "05/08/2018 15:21:29", "content": "Sharing via GitHub is allowed as long as you give a link in a public forum of the competition. That Instacart solution was more than a week before the end (though not much more) IIRC. And it required some effort to implement. (You had to compile Arboretum yourself, for example.) That competition was my second medal, which got me to expert status, so I remember it pretty well. I almost missed out on a medal because of the effect of the shared solution, but in the end I climbed back and learned a lot in the process.
", "rawMarkdown": "Sharing via GitHub is allowed as long as you give a link in a public forum of the competition. [That Instacart solution][1] was more than a week before the end (though not much more) IIRC. And it required some effort to implement. (You had to compile Arboretum yourself, for example.) That competition was my second medal, which got me to expert status, so I remember it pretty well. I almost missed out on a medal because of the effect of the shared solution, but in the end I climbed back and learned a lot in the process.\n\n\n [1]: https://www.kaggle.com/c/instacart-market-basket-analysis/discussion/37697", "votes": null }, { "id": "325580", "postDate": "05/08/2018 15:29:09", "content": "Good to know that sharing via GitHub with a provided link is allowed. And one week before the deadline may not be so disturbing compared to the few hours before the deadline. At least there was some time to react.
", "rawMarkdown": "Good to know that sharing via GitHub with a provided link is allowed. And one week before the deadline may not be so disturbing compared to the few hours before the deadline. At least there was some time to react.", "votes": null }, { "id": "325667", "postDate": "05/08/2018 17:26:39", "content": "What I was saying! @Luca, check the leaderboard if you haven't done it yet. You are in the Bronze, so happy for you :)
", "rawMarkdown": "What I was saying! @Luca, check the leaderboard if you haven't done it yet. You are in the Bronze, so happy for you :)", "votes": null }, { "id": "325669", "postDate": "05/08/2018 17:33:42", "content": "1/2 week(s) would be sufficient for me.
\n\nI think we must bear in mind that it's mainly how knowledge is shared on Kaggle.
", "rawMarkdown": "1/2 week(s) would be sufficient for me. \n\nI think we must bear in mind that it's mainly how knowledge is shared on Kaggle.", "votes": null }, { "id": "325985", "postDate": "05/09/2018 06:37:12", "content": "leader board 0.9821052 , 0.9820619 , 0.981916720
\n\nThey have a very low number of submissions and are identical score. i think that kaggle must take action something.
", "rawMarkdown": "leader board 0.9821052 , 0.9820619 , 0.981916720\n\nThey have a very low number of submissions and are identical score. i think that kaggle must take action something.", "votes": null }, { "id": "326179", "postDate": "05/09/2018 12:15:18", "content": "I hope they will remove some more cheaters, so you can get an acknowledgment too. The point is that it has little sense to have a medal or a rank in such a similar situation :-(
", "rawMarkdown": "I hope they will remove some more cheaters, so you can get an acknowledgment too. The point is that it has little sense to have a medal or a rank in such a similar situation :-(", "votes": null }, { "id": "326323", "postDate": "05/09/2018 15:13:17", "content": "I think Kaggle has already finalized the results so there will be no further changes.
\n\nWell, I managed my time poorly and didn't make my best submission on time. Also, I didn't choose two submissions. As a result, an overfitting model was chosen and the score was further affected by the late kernels. So overall I learned a lot and am happy that participated in this competition.
\n\nEnjoy your deserved medal and hope to see you in other competitions :)
", "rawMarkdown": "I think Kaggle has already finalized the results so there will be no further changes.\n\nWell, I managed my time poorly and didn't make my best submission on time. Also, I didn't choose two submissions. As a result, an overfitting model was chosen and the score was further affected by the late kernels. So overall I learned a lot and am happy that participated in this competition.\n\nEnjoy your deserved medal and hope to see you in other competitions :)", "votes": null }, { "id": "326326", "postDate": "05/09/2018 15:16:14", "content": "Also there are 111 teams with exactly 0.9820619 private score. Wondering what are the chances of all of them being genuine. Not sure on what basis some of the submissions are removed. I had the option of doing a \"rogue\" submission as well but decided against it. Maybe I should go for it next time
", "rawMarkdown": "Also there are 111 teams with exactly 0.9820619 private score. Wondering what are the chances of all of them being genuine. Not sure on what basis some of the submissions are removed. I had the option of doing a \"rogue\" submission as well but decided against it. Maybe I should go for it next time", "votes": null }, { "id": "326328", "postDate": "05/09/2018 15:17:42", "content": "As the rules allow to use publicly shared kernels' results, those scores are not a violation of rules. So Kaggle won't do anything about it. It's very unethical but sadly we can't do anything about it. We can just hope that the rules will improve over time.
", "rawMarkdown": "As the rules allow to use publicly shared kernels' results, those scores are not a violation of rules. So Kaggle won't do anything about it. It's very unethical but sadly we can't do anything about it. We can just hope that the rules will improve over time.", "votes": null }, { "id": "326332", "postDate": "05/09/2018 15:20:23", "content": "@Satya, I wouldn't do it and honestly, I was thinking that if I were in the place of those people (at least those who have only several submissions), I would voluntarily leave the leaderboard.
", "rawMarkdown": "Satya, I wouldn't do it and honestly, I was thinking that if I were in the place of those people (at least those who have only several submissions), I would voluntarily leave the leaderboard.", "votes": null }, { "id": "326335", "postDate": "05/09/2018 15:23:26", "content": "Actually, it happened to me, too, to mess submissions up or simply to decide for the completely wrong ones. I guess it is part of the game and it happens because of high pressure in the last part of the competition. With some more competition experience, you will find that it is helpful having a submission strategy (for instance deciding for the highest leaderboard score and the highest local cv submission which, in my experience, never proved to be a bad choice) See you in the next competitions :-)
", "rawMarkdown": "Actually, it happened to me, too, to mess submissions up or simply to decide for the completely wrong ones. I guess it is part of the game and it happens because of high pressure in the last part of the competition. With some more competition experience, you will find that it is helpful having a submission strategy (for instance deciding for the highest leaderboard score and the highest local cv submission which, in my experience, never proved to be a bad choice) See you in the next competitions :-)", "votes": null }, { "id": "326343", "postDate": "05/09/2018 15:30:23", "content": "I may use that technique to choose submissions next time. I knew I had a model that was generalizing better than the others but I put so much pressure on the last days that I was exhausted. There was no mind left to make a choice or do the ensemble submission on time :)\nLessons learned and see you...
", "rawMarkdown": "I may use that technique to choose submissions next time. I knew I had a model that was generalizing better than the others but I put so much pressure on the last days that I was exhausted. There was no mind left to make a choice or do the ensemble submission on time :)\nLessons learned and see you...", "votes": null }, { "id": "326569", "postDate": "05/09/2018 23:45:58", "content": "I know that this does not violate the rules. so more sad :( \nNext time I will try harder and get medals. Fighting !!!
", "rawMarkdown": "I know that this does not violate the rules. so more sad :( \nNext time I will try harder and get medals. Fighting !!!", "votes": null }, { "id": "327167", "postDate": "05/11/2018 00:03:03", "content": "@Eric, it is brave for you to acknowledge you submit kernel file, applause for you! However, as you said, 'Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...'. If you really say it from heart, may I ask:1). Why you submit kernel file? 2). Even you are just curious to submit the kernel file, why you select it as final submission? Actually, you can learn almost nothing from just downloading a kernel file and submitting it, you did this because you wanted a medal as well...Now I think yo could understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file.
", "rawMarkdown": "Eric, it is brave for you to acknowledge you submit kernel file, applause for you! However, as you said, 'Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...'. If you really say it from heart, may I ask:1). Why you submit kernel file? 2). Even you are just curious to submit the kernel file, why you select it as final submission? Actually, you can learn almost nothing from just downloading a kernel file and submitting it, you did this because you wanted a medal as well...Now I think yo could understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file.", "votes": null }, { "id": "327219", "postDate": "05/11/2018 03:27:31", "content": "Thanks Snorlax, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. \nConcerning your questions, I submitted this kernel as I did when I took over other shared kernels to run them and see how it performed. I honnestly though everyone would do the same and I would just compete with equal arms with other. And concerning the final submission, to be honnest, I did not do on purpose to use it as a final submission. In the last day, I was working like a dog on this as the deadling was coming closer and did not have time to think which should be my final submission. I left it unchecked and the system took my two best submissions....
\n\nBut again, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. In the future, I will be much more attentive to ths king of feature to avoid hangering other nice kagglers competitors.
\n\nLike eveyrone in this forum, I believe in true competition and I am also in favor of banning kernel submitted the last day to avoid this kind of problem.
", "rawMarkdown": "Thanks Snorlax, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. \nConcerning your questions, I submitted this kernel as I did when I took over other shared kernels to run them and see how it performed. I honnestly though everyone would do the same and I would just compete with equal arms with other. And concerning the final submission, to be honnest, I did not do on purpose to use it as a final submission. In the last day, I was working like a dog on this as the deadling was coming closer and did not have time to think which should be my final submission. I left it unchecked and the system took my two best submissions....\n\nBut again, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. In the future, I will be much more attentive to ths king of feature to avoid hangering other nice kagglers competitors.\n\nLike eveyrone in this forum, I believe in true competition and I am also in favor of banning kernel submitted the last day to avoid this kind of problem.", "votes": null } ], "comments": [ { "id": 324883, "author_name": "richinmind", "author_url": "", "post_date": "05/08/2018 01:11:16", "content": "This is a shame!
", "votes": null, "replies": [] }, { "id": 324893, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 01:22:42", "content": "And this is just the copies. Imagine how many people blended their results with those last minute kernels. Lesson learnt, if I participate in a Kaggle competition again, I need to wait for the last minute kernels and blend my work with them.
", "votes": null, "replies": [ { "id": 324906, "author_name": "laevatein", "author_url": "", "post_date": "05/08/2018 01:27:53", "content": "Exactly, that's totally meaningless.
", "votes": null, "replies": [] }, { "id": 324979, "author_name": "liuhdsgoal", "author_url": "", "post_date": "05/08/2018 02:31:31", "content": "Just wait to last day, 'do some blendings, drink a coffee, win medals.' from @Wenjie Bai
", "votes": null, "replies": [] } ] }, { "id": 324905, "author_name": "eigenvector", "author_url": "", "post_date": "05/08/2018 01:27:40", "content": "It's ironic that a Fraud Detection Challenge had such rampant \"fraud\" in the submissions.
", "votes": null, "replies": [] }, { "id": 324965, "author_name": "liuhdsgoal", "author_url": "", "post_date": "05/08/2018 02:19:34", "content": "It's so sad to see someone work hard to get 09800,and only a sharing of 09811 kernel ruins all
", "votes": null, "replies": [] }, { "id": 324970, "author_name": "wenjiebai", "author_url": "", "post_date": "05/08/2018 02:26:17", "content": "Next time maybe we should just wait for the last day, dowload some submissions, do some blendings, drink a coffee, win medals. Why would we pay months of time/effort to achieve nothing? Ironic.
", "votes": null, "replies": [ { "id": 324974, "author_name": "liuhdsgoal", "author_url": "", "post_date": "05/08/2018 02:29:10", "content": "all you need is to stay night for the last day, and wait for last minute to blend some kernels.
", "votes": null, "replies": [] }, { "id": 325463, "author_name": "dicksonchin93", "author_url": "", "post_date": "05/08/2018 12:44:58", "content": "just submit there isn’t a need to blend to get silver anyways why any extra effort ;)
", "votes": null, "replies": [] } ] }, { "id": 324995, "author_name": "aharless", "author_url": "", "post_date": "05/08/2018 02:57:31", "content": "I'm going to repeat a suggestion I made here: submissions from public kernels should be disabled during the last week. (You could still fork public kernels and submit them privately, but people wouldn't be able to create public kernels with verified scores, so this kind of rule would tend to short-circuit the last minute spoiler process.)
", "votes": null, "replies": [ { "id": 325017, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 03:23:06", "content": "That is an excellent suggestion and I vote for it! It will be the best solution.
", "votes": null, "replies": [] }, { "id": 325126, "author_name": "aljaz91", "author_url": "", "post_date": "05/08/2018 06:36:14", "content": "That's part of the solution, but not entirely - if I remember correctly, the spoiler in Instacart Market Basket competition was shared via github. And the impact was huge, I had been in the top silver range but at the end could barely squeeze out a bronze.
", "votes": null, "replies": [] }, { "id": 325518, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 13:48:56", "content": "@Aljaž, that's awful! Sharing via GitHub is even against the rules. Weren't they disqualified?
", "votes": null, "replies": [] }, { "id": 325577, "author_name": "aharless", "author_url": "", "post_date": "05/08/2018 15:21:29", "content": "Sharing via GitHub is allowed as long as you give a link in a public forum of the competition. That Instacart solution was more than a week before the end (though not much more) IIRC. And it required some effort to implement. (You had to compile Arboretum yourself, for example.) That competition was my second medal, which got me to expert status, so I remember it pretty well. I almost missed out on a medal because of the effect of the shared solution, but in the end I climbed back and learned a lot in the process.
", "votes": null, "replies": [] }, { "id": 325580, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 15:29:09", "content": "Good to know that sharing via GitHub with a provided link is allowed. And one week before the deadline may not be so disturbing compared to the few hours before the deadline. At least there was some time to react.
", "votes": null, "replies": [] } ] }, { "id": 325005, "author_name": "reachkishore", "author_url": "", "post_date": "05/08/2018 03:09:59", "content": "I am pasting the same what I had suggested earlier here
\nMy suggestion to @Kaggle is
\nto make an intermediate dead line like ( one/two month after launch of competition) ,
Till this date kernels can be made public.
\nlater this date all new kernels will become private kernels.
\nThese can be made public post competition.
\nBy doing this Blending can still be be done using past public kernels also it encourages Diverse solutions nearing to competition closure. Thank you
Hope to see it happen in my second competition. If it doesn't happen, I will be more prepared.\nA huge thing learned is that I need to spend more time reading discussions and kernels.
", "votes": null, "replies": [] }, { "id": 325669, "author_name": "titidutarn", "author_url": "", "post_date": "05/08/2018 17:33:42", "content": "1/2 week(s) would be sufficient for me.
\n\nI think we must bear in mind that it's mainly how knowledge is shared on Kaggle.
", "votes": null, "replies": [] } ] }, { "id": 325029, "author_name": "marcuslin", "author_url": "", "post_date": "05/08/2018 03:34:04", "content": "I was not comfortable at the beginning, but after final results announced, Top competitor ranks did not change significantly.\nAnd then I realized, if my model were good enough, It wouldn't affect me at all, so to me, there are still many things to learn.
", "votes": null, "replies": [ { "id": 325035, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 03:43:59", "content": "Totally agree, I know that after some point those public kernels wouldn't heart much. But if submitting exactly the same file as provided by last-minute public kernel results in Silver, there is surely something wrong with it.
\n\nI had 3 models scoring 0.9814 and averaged them, submitted at 5:03 pm. If I was on time, that score would provide us with the 222nd place on the private leaderboard and last minute kernels wouldn't heart us much. Even then, we would have a chance to win silver instead of bronze if not the last-minute kernels. But I am more frustrated not because of my unpunctuality but because of the unfair last minute sharings.
\n\nFrom not being punctual, I learned to be more punctual. From last-minute public sharings, I learned that cheating can be legalized. I don't think that the second learning adds any value to my data science skills.
", "votes": null, "replies": [] } ] }, { "id": 325083, "author_name": "satyapati", "author_url": "", "post_date": "05/08/2018 05:24:55", "content": "Very disheartening. Some learnings but feels like a big waste of effort and money :(
", "votes": null, "replies": [ { "id": 325093, "author_name": "wenjiebai", "author_url": "", "post_date": "05/08/2018 05:44:01", "content": "Totally agree with you, if they use true ability to win me i shall be gald to receive learnings, but what kind of learnings this one is counted for? 5hours per day for 40days, i feel now they are completely waste.
", "votes": null, "replies": [] } ] }, { "id": 325153, "author_name": "izmaylov", "author_url": "", "post_date": "05/08/2018 07:15:48", "content": "I thought till the end , that all this public kernels were overfitted and believed in my solution. It was a mistake(
", "votes": null, "replies": [] }, { "id": 325188, "author_name": "lpachuong", "author_url": "", "post_date": "05/08/2018 07:53:12", "content": "It's not the first time, and it won't be the last time. Unfortunately.
\n\nThough, it is not hard to scan through the duplicated results and removed them except the first submission. You can't have 99.9% similarity if it is not plagiarism.
", "votes": null, "replies": [] }, { "id": 325207, "author_name": "phansoks", "author_url": "", "post_date": "05/08/2018 08:14:40", "content": "I don't think there are any direct copied submissions in medals. I've just tested it and it returns a private score of 0.9819167 which is below yours. But, as you said, it may have been used anyways for blending.
", "votes": null, "replies": [ { "id": 325479, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 13:03:10", "content": "You tested only one of them. Check the scores. Exact same scores have almost no chance of coming from different submission files.
", "votes": null, "replies": [] } ] }, { "id": 325303, "author_name": "ericbenhamou", "author_url": "", "post_date": "05/08/2018 09:39:08", "content": "I completely understand the bitterness, resentment and angerness in this competition. I was not in the bronze medal until the last minute. I was about to quit and I saw this kernel. I decided to submit it. I was in a situation where I had nothing to loose. Should this kernel have overfitted the public score dataset, all the crowds would have been wiped out. It turns out the other way...
\n\nBut at the end of the day, this is just a game and the real thing is to learn.... Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...
\n\nLast advice, it seems that if you are not a top kaggler performer, it is very wise to keep your last day submission in case something happens. I did not follow you and i was tracking closely Wenjie Bai and found that she was very clever and inspiring. At the end of the day, this medal is neither a diploma or real money and we should not put too much emphasis on it!!!
\n\nFinally, the Andy Harless suggestion, no kernel the last day seems wise to me.\nAll the best to you and keep it up with kaggle
", "votes": null, "replies": [ { "id": 325486, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 13:14:47", "content": "I will disagree with you. Not with the part of learning. Yes, I learned a lot during the past month. And no one can take the knowledge away from me.
\n\nBut if it wasn't for the medal, you wouldn't take a chance and submit it in the first place. Have you seen how many people participate in research or playground competitions compared to the featured competitions? Yes, a huge difference. And that's because of the medals and prizes!
\n\nKaggle's fame of being a great place for learning is because of the medals and prizes. Let's be honest, we, human beings, like to be praised for our achievements. So let's not be too modest and say that all that matters is the knowledge. First time competing in Kaggle and I had to be in the Bronze, even had a chance for Silver, and that was the motivation which was taken away from me because of the poor construction of the rules.
", "votes": null, "replies": [] }, { "id": 325496, "author_name": "wenjiebai", "author_url": "", "post_date": "05/08/2018 13:25:15", "content": "@Araks, i like your view, and i'd like to regroup guys like you for next competition.
", "votes": null, "replies": [] }, { "id": 325513, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 13:43:22", "content": "Started following you. I have decided to try next competitions solo at least until the merger deadline. We will see how it goes. Will keep in touch with you.
", "votes": null, "replies": [] }, { "id": 327167, "author_name": "wythhh", "author_url": "", "post_date": "05/11/2018 00:03:03", "content": "@Eric, it is brave for you to acknowledge you submit kernel file, applause for you! However, as you said, 'Medals are nice to have but really was is important is to spend hours thinking about the problem and how to model. I learned so much over the last week...'. If you really say it from heart, may I ask:1). Why you submit kernel file? 2). Even you are just curious to submit the kernel file, why you select it as final submission? Actually, you can learn almost nothing from just downloading a kernel file and submitting it, you did this because you wanted a medal as well...Now I think yo could understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file.
", "votes": null, "replies": [] }, { "id": 327219, "author_name": "ericbenhamou", "author_url": "", "post_date": "05/11/2018 03:27:31", "content": "Thanks Snorlax, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. \nConcerning your questions, I submitted this kernel as I did when I took over other shared kernels to run them and see how it performed. I honnestly though everyone would do the same and I would just compete with equal arms with other. And concerning the final submission, to be honnest, I did not do on purpose to use it as a final submission. In the last day, I was working like a dog on this as the deadling was coming closer and did not have time to think which should be my final submission. I left it unchecked and the system took my two best submissions....
\n\nBut again, I fully understand the mood of the people whose medals are 'stolen' from people who just submitted a kernel file and I am sincerely sorry for them. In the future, I will be much more attentive to ths king of feature to avoid hangering other nice kagglers competitors.
\n\nLike eveyrone in this forum, I believe in true competition and I am also in favor of banning kernel submitted the last day to avoid this kind of problem.
", "votes": null, "replies": [] } ] }, { "id": 325471, "author_name": "scirpus", "author_url": "", "post_date": "05/08/2018 12:55:24", "content": "This is by far the worst kaggle has been in my four years - take comfort it will be better next time
\n\nThe only way is up ;)
", "votes": null, "replies": [ { "id": 325526, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 13:54:19", "content": "@Scirpus, great spirit, perfect :)
", "votes": null, "replies": [] } ] }, { "id": 325524, "author_name": "lucamassaron", "author_url": "", "post_date": "05/08/2018 13:54:09", "content": "If it can be of any solace, think of me, just a single place away from the bronze medal :-( A couple of days ago I have been at position 142, actually, after literally throwing money at AWS instances in order to achieve a performing model.
\n\nStoically, I do not say anything about such a surreal competition's ending because I've seen really much far worse situations in 118 competitions I participated in. And I assure you, nothing has ever changed in spite of much protest and discussion. Such is the reality of participating to such competitions. Enjoy what you have learned, really, points and medals are just ephemeral.
", "votes": null, "replies": [ { "id": 325530, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 14:05:22", "content": "@Luca, thanks for being here. I am sure at least one person will be disqualified and you will get your earned medal. Yesterday we were 407th, now I see that we are 406th. So there is a great hope that you will get Bronze.
\n\nThis was a surprise for me as this is my first competition, but I spent quite some time reading other discussions yesterday and realized that it's very normal. Kind of a new motivation for me, to work harder and end in a better place so that this kind of things don't affect my results.
\n\nStill, have hopes for future improvements of the rules.
", "votes": null, "replies": [] }, { "id": 325667, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/08/2018 17:26:39", "content": "What I was saying! @Luca, check the leaderboard if you haven't done it yet. You are in the Bronze, so happy for you :)
", "votes": null, "replies": [] }, { "id": 326179, "author_name": "lucamassaron", "author_url": "", "post_date": "05/09/2018 12:15:18", "content": "I hope they will remove some more cheaters, so you can get an acknowledgment too. The point is that it has little sense to have a medal or a rank in such a similar situation :-(
", "votes": null, "replies": [] }, { "id": 326323, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/09/2018 15:13:17", "content": "I think Kaggle has already finalized the results so there will be no further changes.
\n\nWell, I managed my time poorly and didn't make my best submission on time. Also, I didn't choose two submissions. As a result, an overfitting model was chosen and the score was further affected by the late kernels. So overall I learned a lot and am happy that participated in this competition.
\n\nEnjoy your deserved medal and hope to see you in other competitions :)
", "votes": null, "replies": [] }, { "id": 326335, "author_name": "lucamassaron", "author_url": "", "post_date": "05/09/2018 15:23:26", "content": "Actually, it happened to me, too, to mess submissions up or simply to decide for the completely wrong ones. I guess it is part of the game and it happens because of high pressure in the last part of the competition. With some more competition experience, you will find that it is helpful having a submission strategy (for instance deciding for the highest leaderboard score and the highest local cv submission which, in my experience, never proved to be a bad choice) See you in the next competitions :-)
", "votes": null, "replies": [] }, { "id": 326343, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/09/2018 15:30:23", "content": "I may use that technique to choose submissions next time. I knew I had a model that was generalizing better than the others but I put so much pressure on the last days that I was exhausted. There was no mind left to make a choice or do the ensemble submission on time :)\nLessons learned and see you...
", "votes": null, "replies": [] } ] }, { "id": 325985, "author_name": "chocozzz", "author_url": "", "post_date": "05/09/2018 06:37:12", "content": "leader board 0.9821052 , 0.9820619 , 0.981916720
\n\nThey have a very low number of submissions and are identical score. i think that kaggle must take action something.
", "votes": null, "replies": [ { "id": 326326, "author_name": "satyapati", "author_url": "", "post_date": "05/09/2018 15:16:14", "content": "Also there are 111 teams with exactly 0.9820619 private score. Wondering what are the chances of all of them being genuine. Not sure on what basis some of the submissions are removed. I had the option of doing a \"rogue\" submission as well but decided against it. Maybe I should go for it next time
", "votes": null, "replies": [] }, { "id": 326328, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/09/2018 15:17:42", "content": "As the rules allow to use publicly shared kernels' results, those scores are not a violation of rules. So Kaggle won't do anything about it. It's very unethical but sadly we can't do anything about it. We can just hope that the rules will improve over time.
", "votes": null, "replies": [] }, { "id": 326332, "author_name": "araksstepanyan", "author_url": "", "post_date": "05/09/2018 15:20:23", "content": "@Satya, I wouldn't do it and honestly, I was thinking that if I were in the place of those people (at least those who have only several submissions), I would voluntarily leave the leaderboard.
", "votes": null, "replies": [] }, { "id": 326569, "author_name": "chocozzz", "author_url": "", "post_date": "05/09/2018 23:45:58", "content": "I know that this does not violate the rules. so more sad :( \nNext time I will try harder and get medals. Fighting !!!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "324856": "And we are only 11 places away from the Bronze. I think something needs to be changed!\n\nEdit: 6 places away from the Bronze and looks like this is final.\n\nEdit: Check this out, the leaderboard progression animation: https://www.kaggle.com/inversion/talkingdata-leaderboard-progression/code", "324883": "This is a shame!", "324893": "And this is just the copies. Imagine how many people blended their results with those last minute kernels. Lesson learnt, if I participate in a Kaggle competition again, I need to wait for the last minute kernels and blend my work with them.", "324905": "It's ironic that a Fraud Detection Challenge had such rampant \"fraud\" in the submissions.", "324906": "Exactly, that's totally meaningless.", "324965": "It's so sad to see someone work hard to get 09800,and only a sharing of 09811 kernel ruins all", "324970": "Next time maybe we should just wait for the last day, dowload some submissions, do some blendings, drink a coffee, win medals. Why would we pay months of time/effort to achieve nothing? Ironic.", "324974": "all you need is to stay night for the last day, and wait for last minute to blend some kernels.", "324979": "Just wait to last day, 'do some blendings, drink a coffee, win medals.' from @Wenjie Bai", "324995": "I'm going to repeat a suggestion I made [here][1]: submissions from public kernels should be disabled during the last week. (You could still fork public kernels and submit them privately, but people wouldn't be able to create public kernels with verified scores, so this kind of rule would tend to short-circuit the last minute spoiler process.) \n\n [1]: https://www.kaggle.com/c/talkingdata-adtracking-fraud-detection/discussion/54400#315120", "325005": "I am pasting the same what I had suggested earlier [here][1]