Hi,\nMy theory about @senkin13 being in the 1st spot, given the gap from second and the amount of submissions in compared to others, my guess is that he is going through his top 'IsOutlier' predictions and puts them as -33.12 to see if they make the score better to 100% classify them and make his score better.
\n\nif this is correct - is it true that Private leaderboard DOES NOT contain any card_ids that are in public leaderboard?
\n\nI want to save the guy his time if this is the case =)
\n\n@senkin13 are you doing any other magic?
", "messages": [ { "id": "453852", "postDate": "01/10/2019 22:27:34", "content": "Hi,\nMy theory about @senkin13 being in the 1st spot, given the gap from second and the amount of submissions in compared to others, my guess is that he is going through his top 'IsOutlier' predictions and puts them as -33.12 to see if they make the score better to 100% classify them and make his score better.
\n\nif this is correct - is it true that Private leaderboard DOES NOT contain any card_ids that are in public leaderboard?
\n\nI want to save the guy his time if this is the case =)
\n\n@senkin13 are you doing any other magic?
", "rawMarkdown": "Hi,\nMy theory about @senkin13 being in the 1st spot, given the gap from second and the amount of submissions in compared to others, my guess is that he is going through his top 'IsOutlier' predictions and puts them as -33.12 to see if they make the score better to 100% classify them and make his score better.\n\nif this is correct - is it true that Private leaderboard DOES NOT contain any card_ids that are in public leaderboard?\n\nI want to save the guy his time if this is the case =)\n\n@senkin13 are you doing any other magic?", "votes": null }, { "id": "453943", "postDate": "01/11/2019 01:08:30", "content": "AmirH,thanks your care.I won’t share in competition.But I can say don’t trust LB.
", "rawMarkdown": "AmirH,thanks your care.I won’t share in competition.But I can say don’t trust LB.", "votes": null }, { "id": "454099", "postDate": "01/11/2019 06:07:48", "content": "There is no meaning to just get better public LB socre. Nobody want to do this.
", "rawMarkdown": "There is no meaning to just get better public LB socre. Nobody want to do this.", "votes": null }, { "id": "454114", "postDate": "01/11/2019 06:24:38", "content": "Hi guy, you make a good job.
", "rawMarkdown": "Hi guy, you make a good job.", "votes": null }, { "id": "454173", "postDate": "01/11/2019 07:51:44", "content": "Thank you Wang jianfei, I hope so,but it's too early to say.
", "rawMarkdown": "Thank you Wang jianfei, I hope so,but it's too early to say.", "votes": null }, { "id": "454318", "postDate": "01/11/2019 12:34:16", "content": "Instead of directly asking/questioning someone else's strategy - I say why don't we keep plugging away on this brain teasers :)
\n\nThat being said, I am in absolute confidience that whatever tactics @senkin13 uses in this competition, it is chosen with very valid reasons and I for one can't wait till competition ends to learn about it ^_^
", "rawMarkdown": "Instead of directly asking/questioning someone else's strategy - I say why don't we keep plugging away on this brain teasers :)\n\nThat being said, I am in absolute confidience that whatever tactics @senkin13 uses in this competition, it is chosen with very valid reasons and I for one can't wait till competition ends to learn about it ^_^", "votes": null }, { "id": "454385", "postDate": "01/11/2019 15:04:03", "content": "hmm, sorry for asking, but why people dont want it?
", "rawMarkdown": "hmm, sorry for asking, but why people dont want it?", "votes": null }, { "id": "454735", "postDate": "01/12/2019 05:35:46", "content": "Because private LB is the ultimate target! The score on public LB only represents 30% of testing data. The rest 70% will be measured on private LB. So no matter what you've done on public LB it doesn't contribute to your final ranking, except building a real model successfully predicts the rest 70% testing data.
", "rawMarkdown": "Because private LB is the ultimate target! The score on public LB only represents 30% of testing data. The rest 70% will be measured on private LB. So no matter what you've done on public LB it doesn't contribute to your final ranking, except building a real model successfully predicts the rest 70% testing data.", "votes": null }, { "id": "455978", "postDate": "01/14/2019 22:17:51", "content": "Sometimes its fun! just be sure to submit a generalizable model for the final submission.
", "rawMarkdown": "Sometimes its fun! just be sure to submit a generalizable model for the final submission.", "votes": null }, { "id": "456109", "postDate": "01/15/2019 06:57:38", "content": "me too
", "rawMarkdown": "me too", "votes": null }, { "id": "456232", "postDate": "01/15/2019 11:45:49", "content": "can you share you local cv for your best lb？
", "rawMarkdown": "can you share you local cv for your best lb？", "votes": null } ], "comments": [ { "id": 453943, "author_name": "senkin13", "author_url": "", "post_date": "01/11/2019 01:08:30", "content": "AmirH,thanks your care.I won’t share in competition.But I can say don’t trust LB.
", "votes": null, "replies": [ { "id": 454114, "author_name": "codlife", "author_url": "", "post_date": "01/11/2019 06:24:38", "content": "Hi guy, you make a good job.
", "votes": null, "replies": [] }, { "id": 454173, "author_name": "senkin13", "author_url": "", "post_date": "01/11/2019 07:51:44", "content": "Thank you Wang jianfei, I hope so,but it's too early to say.
", "votes": null, "replies": [] }, { "id": 456232, "author_name": "h4211819", "author_url": "", "post_date": "01/15/2019 11:45:49", "content": "can you share you local cv for your best lb？
", "votes": null, "replies": [] } ] }, { "id": 454099, "author_name": "jiazhuang", "author_url": "", "post_date": "01/11/2019 06:07:48", "content": "There is no meaning to just get better public LB socre. Nobody want to do this.
", "votes": null, "replies": [ { "id": 454385, "author_name": "satansw", "author_url": "", "post_date": "01/11/2019 15:04:03", "content": "hmm, sorry for asking, but why people dont want it?
", "votes": null, "replies": [] }, { "id": 454735, "author_name": "crc1104", "author_url": "", "post_date": "01/12/2019 05:35:46", "content": "Because private LB is the ultimate target! The score on public LB only represents 30% of testing data. The rest 70% will be measured on private LB. So no matter what you've done on public LB it doesn't contribute to your final ranking, except building a real model successfully predicts the rest 70% testing data.
", "votes": null, "replies": [] }, { "id": 455978, "author_name": "morenoh149", "author_url": "", "post_date": "01/14/2019 22:17:51", "content": "Sometimes its fun! just be sure to submit a generalizable model for the final submission.
", "votes": null, "replies": [] } ] }, { "id": 454318, "author_name": "yifanxie", "author_url": "", "post_date": "01/11/2019 12:34:16", "content": "Instead of directly asking/questioning someone else's strategy - I say why don't we keep plugging away on this brain teasers :)
\n\nThat being said, I am in absolute confidience that whatever tactics @senkin13 uses in this competition, it is chosen with very valid reasons and I for one can't wait till competition ends to learn about it ^_^
", "votes": null, "replies": [ { "id": 456109, "author_name": "jiazhuang", "author_url": "", "post_date": "01/15/2019 06:57:38", "content": "me too
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "453852": "Hi,\nMy theory about @senkin13 being in the 1st spot, given the gap from second and the amount of submissions in compared to others, my guess is that he is going through his top 'IsOutlier' predictions and puts them as -33.12 to see if they make the score better to 100% classify them and make his score better.\n\nif this is correct - is it true that Private leaderboard DOES NOT contain any card_ids that are in public leaderboard?\n\nI want to save the guy his time if this is the case =)\n\n@senkin13 are you doing any other magic?", "453943": "AmirH,thanks your care.I won’t share in competition.But I can say don’t trust LB.", "454099": "There is no meaning to just get better public LB socre. Nobody want to do this.", "454114": "Hi guy, you make a good job.", "454173": "Thank you Wang jianfei, I hope so,but it's too early to say.", "454318": "Instead of directly asking/questioning someone else's strategy - I say why don't we keep plugging away on this brain teasers :)\n\nThat being said, I am in absolute confidience that whatever tactics @senkin13 uses in this competition, it is chosen with very valid reasons and I for one can't wait till competition ends to learn about it ^_^", "454385": "hmm, sorry for asking, but why people dont want it?", "454735": "Because private LB is the ultimate target! The score on public LB only represents 30% of testing data. The rest 70% will be measured on private LB. So no matter what you've done on public LB it doesn't contribute to your final ranking, except building a real model successfully predicts the rest 70% testing data.", "455978": "Sometimes its fun! just be sure to submit a generalizable model for the final submission.", "456109": "me too", "456232": "can you share you local cv for your best lb？" }, "source": "meta" }