hi,@William Cukierski,I read all the threads here this weekend,and found a little strange public LB behavior here,here is the summary:
\nAs discussed,if I did not wrongly understand,the 0.997 score means [0.99651 0.99749] , if someone got the 0.997 score by a 0.99651 submission,and then,after his hard work, he improved to 0.99749,his Public Score and his position will not go up.
\nI think public/private LB is a great part of Kaggle competition,we need local CV and public LB score to check our models.
\n
\nSome suggestions:
\n1.Perhaps,we need four digits score like 0.9969,in this competition.To prevent probing public LB,we can use 0.9972-0.9974-0.9976 ,that's to say,use the score with precision 0.0002 or 0.0003 or even 0.0005.I think 0.0002 do mean something here.
\n2.If you don't want to increase to four digits,please let the public LB can be sorted correcttly to reflect the real scores.
Thanks for your consideration.
", "messages": [ { "id": 215223, "postDate": "2017-08-20T14:53:58.130Z", "content": "hi,@William Cukierski,I read all the threads here this weekend,and found a little strange public LB behavior here,here is the summary:
\nAs discussed,if I did not wrongly understand,the 0.997 score means [0.99651 0.99749] , if someone got the 0.997 score by a 0.99651 submission,and then,after his hard work, he improved to 0.99749,his Public Score and his position will not go up.
\nI think public/private LB is a great part of Kaggle competition,we need local CV and public LB score to check our models.
\n
\nSome suggestions:
\n1.Perhaps,we need four digits score like 0.9969,in this competition.To prevent probing public LB,we can use 0.9972-0.9974-0.9976 ,that's to say,use the score with precision 0.0002 or 0.0003 or even 0.0005.I think 0.0002 do mean something here.
\n2.If you don't want to increase to four digits,please let the public LB can be sorted correcttly to reflect the real scores.
Thanks for your consideration.
", "rawMarkdown": "hi,@William Cukierski,I read all the threads here this weekend,and found a little strange public LB behavior here,here is the summary:In trying to overcome LB probing (which is possible when there is several hundreds of test samples) Kaggle admins made LB more or less useless in competition with 100000 test samples (245 billion pixels in total. Try to probe that!).
", "rawMarkdown": "In trying to overcome LB probing (which is possible when there is several hundreds of test samples) Kaggle admins made LB more or less useless in competition with 100000 test samples (245 billion pixels in total. Try to probe that!). ", "votes": 13 }, { "id": 215481, "postDate": "2017-08-21T20:32:43.643Z", "content": "Yup, it seems that our public LB is not accessible any longer in the source LB files. Furthermore, we now know for a fact that LB scores are rounded, and not truncated as previously claimed. I also doubt that the relative improvements within third digit after the decimal are reflected on the LB.
", "rawMarkdown": "Yup, it seems that our public LB is not accessible any longer in the source LB files. Furthermore, we now know for a fact that LB scores are rounded, and not truncated as previously claimed. I also doubt that the relative improvements within third digit after the decimal are reflected on the LB.", "votes": 3 }, { "id": 215331, "postDate": "2017-08-21T07:31:30.650Z", "content": "It seems we can't even see our public LB any longer.
", "rawMarkdown": "It seems we can't even see our public LB any longer.", "votes": 4 }, { "id": 215885, "postDate": "2017-08-23T14:53:29.593Z", "content": "lb_score *= 100 ?
", "rawMarkdown": "lb_score *= 100 ?", "votes": 1 }, { "id": 215276, "postDate": "2017-08-20T21:23:09.547Z", "content": "Hi @beyond,
\n\nI'm not the admin for this competition, but I can say that the position on the LB does reflect the scores out to full precision. In other words, improving a score from 0.99651 to 0.99749 will change a the LB position if there was someone between those scores. If there was no one between those scores, there will be no corresponding improvement in LB placement.
", "rawMarkdown": "Hi @beyond,\n\nI'm not the admin for this competition, but I can say that the position on the LB *does* reflect the scores out to full precision. In other words, improving a score from 0.99651 to 0.99749 will change a the LB position if there was someone between those scores. If there was no one between those scores, there will be no corresponding improvement in LB placement.", "votes": -9, "replies": [ { "id": 215278, "postDate": "2017-08-20T22:15:14.047Z", "content": "Ehmmm, not sure about that. From checking source code of the HTML, I went from previous public score of .0.99552 (which placed in the 0.996 GRP) to 0.99625 with no change in lb position.
", "rawMarkdown": "Ehmmm, not sure about that. From checking source code of the HTML, I went from previous public score of .0.99552 (which placed in the 0.996 GRP) to 0.99625 with no change in lb position.", "votes": 16 }, { "id": 215279, "postDate": "2017-08-20T22:32:39.007Z", "content": "inversion,thanks for your reply
\nAs reported by others,someone went from 0.996 to 0.997 at the first time ,then his position will change,and the position is determined by the real score,there is no problem here.
\nAnd then,if he change the score in [0.99651 0.99749],his posision will not change as discussed in other threads by other competitors here: https://www.kaggle.com/c/carvana-image-masking-challenge/discussion/37137#211980.
\nAnd we can have a look at the public LB,we can find that there is no small UP position change.If someone's position change,the change will be a big jump .
inversion, I really doubt that. By looking at the HTML source, my recent submission was scored at 0.9959, my previous at 0.9955. No change in LB position. With currently almost 100 teams scored at 0.996 I'm quite sure someone was between the two scores.
", "rawMarkdown": "inversion, I really doubt that. By looking at the HTML source, my recent submission was scored at 0.9959, my previous at 0.9955. No change in LB position. With currently almost 100 teams scored at 0.996 I'm quite sure someone was between the two scores. \n", "votes": 5 }, { "id": 215350, "postDate": "2017-08-21T09:24:06.390Z", "content": "The probability that 100+ kagglers in the 0.996 band submit the first time and then never improve their results (compared to other 0.996ers) is infinitesimal.
", "rawMarkdown": "The probability that 100+ kagglers in the 0.996 band submit the first time and then never improve their results (compared to other 0.996ers) is infinitesimal. ", "votes": 6 }, { "id": 215396, "postDate": "2017-08-21T12:54:58.073Z", "content": "The feedback is much appreciated! I'll look into it an see what's up.
", "rawMarkdown": "The feedback is much appreciated! I'll look into it an see what's up.", "votes": 5 }, { "id": 215403, "postDate": "2017-08-21T13:46:41.890Z", "content": "From my side I can point out that I have submitted a solution which is better than previous one ( I checked it via sorting in Sub list and html code) but the form told me that it is not better than previous one - and that is obviously wrong
", "rawMarkdown": "From my side I can point out that I have submitted a solution which is better than previous one ( I checked it via sorting in Sub list and html code) but the form told me that it is not better than previous one - and that is obviously wrong", "votes": 3 }, { "id": 215446, "postDate": "2017-08-21T17:44:08.907Z", "content": "It seems that there have been some changes in the scoring today. I also believed that my today score is better than yesterday (via sorting in sub list same as Valeriy) but it says not an improvement.
", "rawMarkdown": "It seems that there have been some changes in the scoring today. I also believed that my today score is better than yesterday (via sorting in sub list same as Valeriy) but it says not an improvement.", "votes": 1 }, { "id": 215608, "postDate": "2017-08-22T09:59:42.803Z", "content": "\n\n\nThe feedback is much appreciated!
\n
@inversion, appreciated? Sad joke. This situation (the reaction of 20 people sharing the same truncated score as the guy at 3rd place) was predicted by me weeks before launching Carvana competition during the discussion regarding scores in Mercedes-Benz and in Web Traffic competitions and the only appreciation were nice thanks for my opponents.
\n\nhttps://www.kaggle.com/c/mercedes-benz-greener-manufacturing/discussion/35307#198647
\n\nhttps://www.kaggle.com/c/web-traffic-time-series-forecasting/discussion/36325#203387
\n\nEDIT: And the reaction of my fans is still the same - downvoting. Thanks, going back to paid prediction.
", "rawMarkdown": ">The feedback is much appreciated!\n\n@inversion, appreciated? Sad joke. This situation (the reaction of 20 people sharing the same truncated score as the guy at 3rd place) was predicted by me weeks before launching Carvana competition during the discussion regarding scores in Mercedes-Benz and in Web Traffic competitions and the only appreciation were nice thanks for my opponents.\n\nhttps://www.kaggle.com/c/mercedes-benz-greener-manufacturing/discussion/35307#198647\n\nhttps://www.kaggle.com/c/web-traffic-time-series-forecasting/discussion/36325#203387\n\nEDIT: And the reaction of my fans is still the same - downvoting. Thanks, going back to paid prediction.\n\n", "votes": 1 }, { "id": 215900, "postDate": "2017-08-23T15:56:50.503Z", "content": "Any progress on this? I find the current situation very demotivating.
", "rawMarkdown": "Any progress on this? I find the current situation very demotivating.", "votes": 10 }, { "id": 215976, "postDate": "2017-08-23T20:31:31.460Z", "content": "Looks like we are totally blind now
", "rawMarkdown": "Looks like we are totally blind now", "votes": 2 }, { "id": 215989, "postDate": "2017-08-23T21:22:02.923Z", "content": "Due to significant number of ground truth mistakes additional digit seems meaningless.\nAs they claim \"the final standings may be different\". This \"different\" can do a disservice.
", "rawMarkdown": "Due to significant number of ground truth mistakes additional digit seems meaningless.\nAs they claim \"the final standings may be different\". This \"different\" can do a disservice." }, { "id": 215991, "postDate": "2017-08-23T21:23:22.623Z", "content": "Just to provide an update - we are still double checking the logic but think we have the problem identified, and expect a fix soon (hopefully in the next few days).
", "rawMarkdown": "Just to provide an update - we are still double checking the logic but think we have the problem identified, and expect a fix soon (hopefully in the next few days).", "votes": 15 }, { "id": 216058, "postDate": "2017-08-24T06:38:04.003Z", "content": "Thanks. So, an official confirmation that the problem is actually there! Good to know, hopefully when it gets fixed the public LB positions will be in the correct order!
", "rawMarkdown": "Thanks. So, an official confirmation that the problem is actually there! Good to know, hopefully when it gets fixed the public LB positions will be in the correct order!", "votes": 4 }, { "id": 216096, "postDate": "2017-08-24T09:14:22.360Z", "content": "@William, My prediction regarding people's reaction on too much truncated LB scores was made in the times when absolutely everybody believed that truncating is made properly and the algorithm sorting LB scores works perfect, so it is still actual. The question is: How many competitions with 20 losers sharing the same truncated score as the guys in the money is needed to realise that such competitions were badly prepared/maintained ones: 3, 5, 10? IMHO: 0.
", "rawMarkdown": "@William, My prediction regarding people's reaction on too much truncated LB scores was made in the times when absolutely everybody believed that truncating is made properly and the algorithm sorting LB scores works perfect, so it is still actual. The question is: How many competitions with 20 losers sharing the same truncated score as the guys in the money is needed to realise that such competitions were badly prepared/maintained ones: 3, 5, 10? IMHO: 0.\n\n" }, { "id": 216794, "postDate": "2017-08-28T06:49:52.103Z", "content": "Hi William, I see the leaderboard is not fixed yet. Sorry for bothering you, but is there anything new?\nThank you.
", "rawMarkdown": "Hi William, I see the leaderboard is not fixed yet. Sorry for bothering you, but is there anything new?\nThank you.", "votes": 3 }, { "id": 216862, "postDate": "2017-08-28T13:09:32.787Z", "content": "Hi William,
\n\nI confirm the current public leaderboard is hardly usable because of low precision of the score and unclear/incorrect sorting. About 50 people now share the same score...
\n\nYou've mentioned in other thread, that a challenge needs to be fun. And the leaderboard is a key functionality and motivation in any challenge. So the information there has to be correct and precise enough.
\n\nAs a new kaggler, I would expect this crucial part to function properly and fixed quickly. \nPlease keep us informed.
\n\nThank you
", "rawMarkdown": "Hi William, \n\nI confirm the current public leaderboard is hardly usable because of low precision of the score and unclear/incorrect sorting. About 50 people now share the same score... \n\nYou've mentioned in other thread, that a challenge needs to be fun. And the leaderboard is a key functionality and motivation in any challenge. So the information there has to be correct and precise enough.\n\nAs a new kaggler, I would expect this **crucial** part to function properly and fixed quickly. \nPlease keep us informed.\n\nThank you\n", "votes": 1 }, { "id": 216884, "postDate": "2017-08-28T14:56:49.453Z", "content": "UPDATE: The problem has been identified, a fix has been developed, and we're working on the final validation of the change. We very much understand this is causing a massive headache for the competitors, and will communicate as soon as the fix is pushed.
", "rawMarkdown": "**UPDATE**: The problem has been identified, a fix has been developed, and we're working on the final validation of the change. We very much understand this is causing a massive headache for the competitors, and will communicate as soon as the fix is pushed.", "votes": 9 }, { "id": 216914, "postDate": "2017-08-28T16:59:40.577Z", "content": "Thanks for the update @inversion. Any estimate on how long it may take before the changes have been implemented? The problem with this competition in particular is that building models takes a long time, and every hour or day that we lose without knowing our true standing is a major delay and setback.
", "rawMarkdown": "Thanks for the update @inversion. Any estimate on how long it may take before the changes have been implemented? The problem with this competition in particular is that building models takes a long time, and every hour or day that we lose without knowing our true standing is a major delay and setback." }, { "id": 216919, "postDate": "2017-08-28T17:13:36.520Z", "content": "Hi Bojan - Tomorrow is the most probably timing. One of the key individuals involved in this change had an unexpected out of the office today.
", "rawMarkdown": "Hi Bojan - Tomorrow is the most probably timing. One of the key individuals involved in this change had an unexpected out of the office today.", "votes": 1 }, { "id": 217049, "postDate": "2017-08-29T06:48:38.303Z", "content": "Thank you for this update. Eagerly waiting for the LB to update properly and see if the last 7 days of work were useful or not!
", "rawMarkdown": "Thank you for this update. Eagerly waiting for the LB to update properly and see if the last 7 days of work were useful or not!" }, { "id": 217169, "postDate": "2017-08-29T16:33:57.600Z", "rawMarkdown": "", "votes": 4, "isDeleted": true }, { "id": 217339, "postDate": "2017-08-30T08:14:29.207Z", "content": "@inversion did that individual ever return? I'm worried about them like never before.
", "rawMarkdown": "@inversion did that individual ever return? I'm worried about them like never before.", "votes": 3 }, { "id": 217359, "postDate": "2017-08-30T10:03:08.953Z", "content": "He is perhaps too shocked after destroying the source code LB
", "rawMarkdown": "He is perhaps too shocked after destroying the source code LB", "votes": -2 }, { "id": 217707, "postDate": "2017-08-31T17:38:27.207Z", "content": "Hey, since there is no change yet, could we pleave have the html source code score back for the meantime?
", "rawMarkdown": "Hey, since there is no change yet, could we pleave have the html source code score back for the meantime? ", "votes": 1 } ] }, { "id": 217062, "postDate": "2017-08-29T08:39:51.653Z", "rawMarkdown": "", "votes": -2 }, { "id": 217055, "postDate": "2017-08-29T07:42:30.540Z", "content": "right
", "rawMarkdown": "right" }, { "id": 216945, "postDate": "2017-08-28T18:37:29.237Z", "content": "How are people knowing what LB score they have beyond the first few decimal points?
", "rawMarkdown": "How are people knowing what LB score they have beyond the first few decimal points?", "replies": [ { "id": 216974, "postDate": "2017-08-28T20:27:28.873Z", "content": "Until recently, you could find it out by inspecting Kaggle's guts. You've missed your chance though :)
\n\nhttps://www.kaggle.com/c/carvana-image-masking-challenge/discussion/38301
", "rawMarkdown": "Until recently, you could find it out by inspecting Kaggle's guts. You've missed your chance though :)\n\nhttps://www.kaggle.com/c/carvana-image-masking-challenge/discussion/38301" }, { "id": 217045, "postDate": "2017-08-29T06:40:00.760Z", "content": "Let's hope they will get back to us soon..
", "rawMarkdown": "Let's hope they will get back to us soon.." } ] }, { "id": 215605, "postDate": "2017-08-22T09:41:13.583Z", "content": "right
", "rawMarkdown": "right" } ], "comments": [ { "id": 215336, "author_name": "Cut Onion", "author_url": "", "post_date": "2017-08-21T07:58:47.443000", "content": "In trying to overcome LB probing (which is possible when there is several hundreds of test samples) Kaggle admins made LB more or less useless in competition with 100000 test samples (245 billion pixels in total. Try to probe that!).
", "votes": 13, "replies": [] }, { "id": 215481, "author_name": "Bojan Tunguz", "author_url": "", "post_date": "2017-08-21T20:32:43.643000", "content": "Yup, it seems that our public LB is not accessible any longer in the source LB files. Furthermore, we now know for a fact that LB scores are rounded, and not truncated as previously claimed. I also doubt that the relative improvements within third digit after the decimal are reflected on the LB.
", "votes": 3, "replies": [] }, { "id": 215331, "author_name": "Harun Gunaydin", "author_url": "", "post_date": "2017-08-21T07:31:30.650000", "content": "It seems we can't even see our public LB any longer.
", "votes": 4, "replies": [] }, { "id": 215885, "author_name": "HariSeldon", "author_url": "", "post_date": "2017-08-23T14:53:29.593000", "content": "lb_score *= 100 ?
", "votes": 1, "replies": [] }, { "id": 215276, "author_name": "inversion", "author_url": "", "post_date": "2017-08-20T21:23:09.547000", "content": "Hi @beyond,
\n\nI'm not the admin for this competition, but I can say that the position on the LB does reflect the scores out to full precision. In other words, improving a score from 0.99651 to 0.99749 will change a the LB position if there was someone between those scores. If there was no one between those scores, there will be no corresponding improvement in LB placement.
", "votes": -9, "replies": [ { "id": 215278, "author_name": "Darragh", "author_url": "", "post_date": "2017-08-20T22:15:14.047000", "content": "Ehmmm, not sure about that. From checking source code of the HTML, I went from previous public score of .0.99552 (which placed in the 0.996 GRP) to 0.99625 with no change in lb position.
", "votes": 16, "replies": [] }, { "id": 215279, "author_name": "beyond", "author_url": "", "post_date": "2017-08-20T22:32:39.007000", "content": "inversion,thanks for your reply
\nAs reported by others,someone went from 0.996 to 0.997 at the first time ,then his position will change,and the position is determined by the real score,there is no problem here.
\nAnd then,if he change the score in [0.99651 0.99749],his posision will not change as discussed in other threads by other competitors here: https://www.kaggle.com/c/carvana-image-masking-challenge/discussion/37137#211980.
\nAnd we can have a look at the public LB,we can find that there is no small UP position change.If someone's position change,the change will be a big jump .
inversion, I really doubt that. By looking at the HTML source, my recent submission was scored at 0.9959, my previous at 0.9955. No change in LB position. With currently almost 100 teams scored at 0.996 I'm quite sure someone was between the two scores.
", "votes": 5, "replies": [] }, { "id": 215350, "author_name": "Selim Seferbekov", "author_url": "", "post_date": "2017-08-21T09:24:06.390000", "content": "The probability that 100+ kagglers in the 0.996 band submit the first time and then never improve their results (compared to other 0.996ers) is infinitesimal.
", "votes": 6, "replies": [] }, { "id": 215396, "author_name": "inversion", "author_url": "", "post_date": "2017-08-21T12:54:58.073000", "content": "The feedback is much appreciated! I'll look into it an see what's up.
", "votes": 5, "replies": [] }, { "id": 215403, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2017-08-21T13:46:41.890000", "content": "From my side I can point out that I have submitted a solution which is better than previous one ( I checked it via sorting in Sub list and html code) but the form told me that it is not better than previous one - and that is obviously wrong
", "votes": 3, "replies": [] }, { "id": 215446, "author_name": "Kaggoo", "author_url": "", "post_date": "2017-08-21T17:44:08.907000", "content": "It seems that there have been some changes in the scoring today. I also believed that my today score is better than yesterday (via sorting in sub list same as Valeriy) but it says not an improvement.
", "votes": 1, "replies": [] }, { "id": 215608, "author_name": "Grzegorz Sionkowski", "author_url": "", "post_date": "2017-08-22T09:59:42.803000", "content": "\n\n\nThe feedback is much appreciated!
\n
@inversion, appreciated? Sad joke. This situation (the reaction of 20 people sharing the same truncated score as the guy at 3rd place) was predicted by me weeks before launching Carvana competition during the discussion regarding scores in Mercedes-Benz and in Web Traffic competitions and the only appreciation were nice thanks for my opponents.
\n\nhttps://www.kaggle.com/c/mercedes-benz-greener-manufacturing/discussion/35307#198647
\n\nhttps://www.kaggle.com/c/web-traffic-time-series-forecasting/discussion/36325#203387
\n\nEDIT: And the reaction of my fans is still the same - downvoting. Thanks, going back to paid prediction.
", "votes": 1, "replies": [] }, { "id": 215900, "author_name": "Nikolay Shebanov", "author_url": "", "post_date": "2017-08-23T15:56:50.503000", "content": "Any progress on this? I find the current situation very demotivating.
", "votes": 10, "replies": [] }, { "id": 215976, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2017-08-23T20:31:31.460000", "content": "Looks like we are totally blind now
", "votes": 2, "replies": [] }, { "id": 215989, "author_name": "Integral", "author_url": "", "post_date": "2017-08-23T21:22:02.923000", "content": "Due to significant number of ground truth mistakes additional digit seems meaningless.\nAs they claim \"the final standings may be different\". This \"different\" can do a disservice.
", "votes": 0, "replies": [] }, { "id": 215991, "author_name": "Will Cukierski", "author_url": "", "post_date": "2017-08-23T21:23:22.623000", "content": "Just to provide an update - we are still double checking the logic but think we have the problem identified, and expect a fix soon (hopefully in the next few days).
", "votes": 15, "replies": [] }, { "id": 216058, "author_name": "Davide Boschetto", "author_url": "", "post_date": "2017-08-24T06:38:04.003000", "content": "Thanks. So, an official confirmation that the problem is actually there! Good to know, hopefully when it gets fixed the public LB positions will be in the correct order!
", "votes": 4, "replies": [] }, { "id": 216096, "author_name": "Grzegorz Sionkowski", "author_url": "", "post_date": "2017-08-24T09:14:22.360000", "content": "@William, My prediction regarding people's reaction on too much truncated LB scores was made in the times when absolutely everybody believed that truncating is made properly and the algorithm sorting LB scores works perfect, so it is still actual. The question is: How many competitions with 20 losers sharing the same truncated score as the guys in the money is needed to realise that such competitions were badly prepared/maintained ones: 3, 5, 10? IMHO: 0.
", "votes": 0, "replies": [] }, { "id": 216794, "author_name": "Davide Boschetto", "author_url": "", "post_date": "2017-08-28T06:49:52.103000", "content": "Hi William, I see the leaderboard is not fixed yet. Sorry for bothering you, but is there anything new?\nThank you.
", "votes": 3, "replies": [] }, { "id": 216862, "author_name": "Roman Semenyk", "author_url": "", "post_date": "2017-08-28T13:09:32.787000", "content": "Hi William,
\n\nI confirm the current public leaderboard is hardly usable because of low precision of the score and unclear/incorrect sorting. About 50 people now share the same score...
\n\nYou've mentioned in other thread, that a challenge needs to be fun. And the leaderboard is a key functionality and motivation in any challenge. So the information there has to be correct and precise enough.
\n\nAs a new kaggler, I would expect this crucial part to function properly and fixed quickly. \nPlease keep us informed.
\n\nThank you
", "votes": 1, "replies": [] }, { "id": 216884, "author_name": "inversion", "author_url": "", "post_date": "2017-08-28T14:56:49.453000", "content": "UPDATE: The problem has been identified, a fix has been developed, and we're working on the final validation of the change. We very much understand this is causing a massive headache for the competitors, and will communicate as soon as the fix is pushed.
", "votes": 9, "replies": [] }, { "id": 216914, "author_name": "Bojan Tunguz", "author_url": "", "post_date": "2017-08-28T16:59:40.577000", "content": "Thanks for the update @inversion. Any estimate on how long it may take before the changes have been implemented? The problem with this competition in particular is that building models takes a long time, and every hour or day that we lose without knowing our true standing is a major delay and setback.
", "votes": 0, "replies": [] }, { "id": 216919, "author_name": "inversion", "author_url": "", "post_date": "2017-08-28T17:13:36.520000", "content": "Hi Bojan - Tomorrow is the most probably timing. One of the key individuals involved in this change had an unexpected out of the office today.
", "votes": 1, "replies": [] }, { "id": 217049, "author_name": "Davide Boschetto", "author_url": "", "post_date": "2017-08-29T06:48:38.303000", "content": "Thank you for this update. Eagerly waiting for the LB to update properly and see if the last 7 days of work were useful or not!
", "votes": 0, "replies": [] }, { "id": 217169, "author_name": "", "author_url": "", "post_date": "2017-08-29T16:33:57.600000", "content": "", "votes": 4, "replies": [] }, { "id": 217339, "author_name": "Nikolay Shebanov", "author_url": "", "post_date": "2017-08-30T08:14:29.207000", "content": "@inversion did that individual ever return? I'm worried about them like never before.
", "votes": 3, "replies": [] }, { "id": 217359, "author_name": "ValeriyBabushkin", "author_url": "", "post_date": "2017-08-30T10:03:08.953000", "content": "He is perhaps too shocked after destroying the source code LB
", "votes": -2, "replies": [] }, { "id": 217707, "author_name": "Charles Jansen", "author_url": "", "post_date": "2017-08-31T17:38:27.207000", "content": "Hey, since there is no change yet, could we pleave have the html source code score back for the meantime?
", "votes": 1, "replies": [] } ] }, { "id": 217062, "author_name": "Grzegorz Sionkowski", "author_url": "", "post_date": "2017-08-29T08:39:51.653000", "content": "", "votes": -2, "replies": [] }, { "id": 217055, "author_name": "jia xu", "author_url": "", "post_date": "2017-08-29T07:42:30.540000", "content": "right
", "votes": 0, "replies": [] }, { "id": 216945, "author_name": "mshliselberg", "author_url": "", "post_date": "2017-08-28T18:37:29.237000", "content": "How are people knowing what LB score they have beyond the first few decimal points?
", "votes": 0, "replies": [ { "id": 216974, "author_name": "Nikolay Shebanov", "author_url": "", "post_date": "2017-08-28T20:27:28.873000", "content": "Until recently, you could find it out by inspecting Kaggle's guts. You've missed your chance though :)
\n\nhttps://www.kaggle.com/c/carvana-image-masking-challenge/discussion/38301
", "votes": 0, "replies": [] }, { "id": 217045, "author_name": "Davide Boschetto", "author_url": "", "post_date": "2017-08-29T06:40:00.760000", "content": "Let's hope they will get back to us soon..
", "votes": 0, "replies": [] } ] }, { "id": 215605, "author_name": "faronFeng", "author_url": "", "post_date": "2017-08-22T09:41:13.583000", "content": "right
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "215223": "hi,@William Cukierski,I read all the threads here this weekend,and found a little strange public LB behavior here,here is the summary: