{
  "id": 297713,
  "title": "Shake up estimation",
  "url": "/competitions/sartorius-cell-instance-segmentation/discussion/297713",
  "author_name": "",
  "post_date": "2021-12-29T01:37:52.120601100Z",
  "votes": 15,
  "comment_count": 19,
  "views": 0,
  "content": "<p>Will there be a big shake up or just little, what do you think?</p>",
  "messages": [
    {
      "id": "1631958",
      "postDate": "12/29/2021 01:37:52",
      "content": "<p>Will there be a big shake up or just little, what do you think?</p>",
      "rawMarkdown": "Will there be a big shake up or just little, what do you think?",
      "votes": null
    },
    {
      "id": "1631961",
      "postDate": "12/29/2021 01:46:41",
      "content": "<p>The private testset is small. <br>\nquite big shake up for some teams, but not the top ones I think.</p>",
      "rawMarkdown": "The private testset is small. \nquite big shake up for some teams, but not the top ones I think.",
      "votes": null
    },
    {
      "id": "1631979",
      "postDate": "12/29/2021 03:08:01",
      "content": "<p>If one overfit the public leaderboard, it will be much easier for him to reach higher LB. So I think the chance of shaking up is even.</p>",
      "rawMarkdown": "If one overfit the public leaderboard, it will be much easier for him to reach higher LB. So I think the chance of shaking up is even.",
      "votes": null
    },
    {
      "id": "1632008",
      "postDate": "12/29/2021 04:13:01",
      "content": "<p>I think a shake of about 0.005 is possible, depending on \"cort\" prevalence.</p>",
      "rawMarkdown": "I think a shake of about 0.005 is possible, depending on \"cort\" prevalence.",
      "votes": null
    },
    {
      "id": "1632028",
      "postDate": "12/29/2021 04:58:04",
      "content": "<p>±5-10 in gold zone, ±10-20 silver zone, ±50 in bronze zone</p>",
      "rawMarkdown": "±5-10 in gold zone, ±10-20 silver zone, ±50 in bronze zone",
      "votes": null
    },
    {
      "id": "1632039",
      "postDate": "12/29/2021 05:21:47",
      "content": "<p>I hope no shake.</p>",
      "rawMarkdown": "I hope no shake.",
      "votes": null
    },
    {
      "id": "1632054",
      "postDate": "12/29/2021 05:55:22",
      "content": "<p>This is exactly what I hope too.</p>",
      "rawMarkdown": "This is exactly what I hope too.",
      "votes": null
    },
    {
      "id": "1632120",
      "postDate": "12/29/2021 08:02:36",
      "content": "<p>I think <a href=\"https://www.kaggle.com/takuok\" target=\"_blank\">@takuok</a> and <a href=\"https://www.kaggle.com/tascj0\" target=\"_blank\">@tascj0</a> will stay 1st if they keep their 0.003 lead on the public LB, as I expect a +/- 0.002 variation. <br>\nHowever 0.345 is not that far from 0.343, the rest of the top 5 could change a bit !<br>\nMy assumption is that organizers did a good job splitting the data :)</p>\n<p>Overall the private scores may be lower than public scores as I believe cort cells are over-represented in the public LB (which explains LB &gt; CV). </p>",
      "rawMarkdown": "I think @takuok and @tascj0 will stay 1st if they keep their 0.003 lead on the public LB, as I expect a +/- 0.002 variation. \nHowever 0.345 is not that far from 0.343, the rest of the top 5 could change a bit !\nMy assumption is that organizers did a good job splitting the data :)\n\nOverall the private scores may be lower than public scores as I believe cort cells are over-represented in the public LB (which explains LB > CV).",
      "votes": null
    },
    {
      "id": "1632230",
      "postDate": "12/29/2021 11:36:32",
      "content": "<p>I hope no shakeup with my team :))</p>",
      "rawMarkdown": "I hope no shakeup with my team :))",
      "votes": null
    },
    {
      "id": "1632262",
      "postDate": "12/29/2021 12:36:10",
      "content": "<p>I pray for a shakeup &gt; 0.012~&gt;&lt;~</p>",
      "rawMarkdown": "I pray for a shakeup > 0.012~><~",
      "votes": null
    },
    {
      "id": "1632523",
      "postDate": "12/29/2021 18:21:29",
      "content": "<p>Hoping for a shake up for us =)</p>",
      "rawMarkdown": "Hoping for a shake up for us =)",
      "votes": null
    },
    {
      "id": "1632592",
      "postDate": "12/29/2021 19:58:48",
      "content": "<p>Cant agree more</p>",
      "rawMarkdown": "Cant agree more",
      "votes": null
    },
    {
      "id": "1632645",
      "postDate": "12/29/2021 21:58:18",
      "content": "<p>If there is a shake up, it will most probably be caused by a shift in class prevalence. This could be the case like <a href=\"https://www.kaggle.com/theoviel\" target=\"_blank\">@theoviel</a> mentionned because cort was clearly over-represented in public LB compared to training distribution.</p>",
      "rawMarkdown": "If there is a shake up, it will most probably be caused by a shift in class prevalence. This could be the case like @theoviel mentionned because cort was clearly over-represented in public LB compared to training distribution.",
      "votes": null
    },
    {
      "id": "1632943",
      "postDate": "12/30/2021 08:47:14",
      "content": "<p>wait ,first tell me what is this Shake :&gt;</p>",
      "rawMarkdown": "wait ,first tell me what is this Shake :>",
      "votes": null
    },
    {
      "id": "1633327",
      "postDate": "12/30/2021 17:19:24",
      "content": "<p>It depends on the distribution of the three classes(shsy5y, cort, astro), in the public LB cort may have more which contribute to higher score.And shsy5y and astro is easily being overfitting.Hope a shake up to a higher rank🙏</p>",
      "rawMarkdown": "It depends on the distribution of the three classes(shsy5y, cort, astro), in the public LB cort may have more which contribute to higher score.And shsy5y and astro is easily being overfitting.Hope a shake up to a higher rank🙏",
      "votes": null
    },
    {
      "id": "1633443",
      "postDate": "12/30/2021 19:12:42",
      "content": "<p><a href=\"https://www.kaggle.com/balavashan\" target=\"_blank\">@balavashan</a> Shakeup means a change in rankings when private ranking is calculated</p>",
      "rawMarkdown": "balavashan Shakeup means a change in rankings when private ranking is calculated",
      "votes": null
    },
    {
      "id": "1633607",
      "postDate": "12/31/2021 00:05:39",
      "content": "<p>Thanks god, nearly no shake up. Happy new year!</p>",
      "rawMarkdown": "Thanks god, nearly no shake up. Happy new year!",
      "votes": null
    },
    {
      "id": "1633615",
      "postDate": "12/31/2021 00:21:14",
      "content": "<p>Grats to all the winners, looking forward to your solutions! </p>",
      "rawMarkdown": "Grats to all the winners, looking forward to your solutions!",
      "votes": null
    },
    {
      "id": "1633635",
      "postDate": "12/31/2021 00:36:26",
      "content": "<p>🙁                </p>",
      "rawMarkdown": "🙁",
      "votes": null
    },
    {
      "id": "1633717",
      "postDate": "12/31/2021 02:08:12",
      "content": "<p>Sorry for your shake down, good luck next time!</p>",
      "rawMarkdown": "Sorry for your shake down, good luck next time!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1631961,
      "author_name": "tiandaye",
      "author_url": "",
      "post_date": "12/29/2021 01:46:41",
      "content": "<p>The private testset is small. <br>\nquite big shake up for some teams, but not the top ones I think.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1631979,
          "author_name": "steamedsheep",
          "author_url": "",
          "post_date": "12/29/2021 03:08:01",
          "content": "<p>If one overfit the public leaderboard, it will be much easier for him to reach higher LB. So I think the chance of shaking up is even.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1632008,
      "author_name": "tyaiga",
      "author_url": "",
      "post_date": "12/29/2021 04:13:01",
      "content": "<p>I think a shake of about 0.005 is possible, depending on \"cort\" prevalence.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632028,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "12/29/2021 04:58:04",
      "content": "<p>±5-10 in gold zone, ±10-20 silver zone, ±50 in bronze zone</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632039,
      "author_name": "markunys",
      "author_url": "",
      "post_date": "12/29/2021 05:21:47",
      "content": "<p>I hope no shake.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1632054,
          "author_name": "steamedsheep",
          "author_url": "",
          "post_date": "12/29/2021 05:55:22",
          "content": "<p>This is exactly what I hope too.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1632120,
      "author_name": "theoviel",
      "author_url": "",
      "post_date": "12/29/2021 08:02:36",
      "content": "<p>I think <a href=\"https://www.kaggle.com/takuok\" target=\"_blank\">@takuok</a> and <a href=\"https://www.kaggle.com/tascj0\" target=\"_blank\">@tascj0</a> will stay 1st if they keep their 0.003 lead on the public LB, as I expect a +/- 0.002 variation. <br>\nHowever 0.345 is not that far from 0.343, the rest of the top 5 could change a bit !<br>\nMy assumption is that organizers did a good job splitting the data :)</p>\n<p>Overall the private scores may be lower than public scores as I believe cort cells are over-represented in the public LB (which explains LB &gt; CV). </p>",
      "votes": null,
      "replies": [
        {
          "id": 1632592,
          "author_name": "charonwangg",
          "author_url": "",
          "post_date": "12/29/2021 19:58:48",
          "content": "<p>Cant agree more</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1632230,
      "author_name": "duykhanh99",
      "author_url": "",
      "post_date": "12/29/2021 11:36:32",
      "content": "<p>I hope no shakeup with my team :))</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632262,
      "author_name": "atom1231",
      "author_url": "",
      "post_date": "12/29/2021 12:36:10",
      "content": "<p>I pray for a shakeup &gt; 0.012~&gt;&lt;~</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632523,
      "author_name": "trushk",
      "author_url": "",
      "post_date": "12/29/2021 18:21:29",
      "content": "<p>Hoping for a shake up for us =)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632645,
      "author_name": "alexandrecc",
      "author_url": "",
      "post_date": "12/29/2021 21:58:18",
      "content": "<p>If there is a shake up, it will most probably be caused by a shift in class prevalence. This could be the case like <a href=\"https://www.kaggle.com/theoviel\" target=\"_blank\">@theoviel</a> mentionned because cort was clearly over-represented in public LB compared to training distribution.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1632943,
      "author_name": "balavashan",
      "author_url": "",
      "post_date": "12/30/2021 08:47:14",
      "content": "<p>wait ,first tell me what is this Shake :&gt;</p>",
      "votes": null,
      "replies": [
        {
          "id": 1633443,
          "author_name": "keagle",
          "author_url": "",
          "post_date": "12/30/2021 19:12:42",
          "content": "<p><a href=\"https://www.kaggle.com/balavashan\" target=\"_blank\">@balavashan</a> Shakeup means a change in rankings when private ranking is calculated</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1633327,
      "author_name": "manyuli",
      "author_url": "",
      "post_date": "12/30/2021 17:19:24",
      "content": "<p>It depends on the distribution of the three classes(shsy5y, cort, astro), in the public LB cort may have more which contribute to higher score.And shsy5y and astro is easily being overfitting.Hope a shake up to a higher rank🙏</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1633607,
      "author_name": "steamedsheep",
      "author_url": "",
      "post_date": "12/31/2021 00:05:39",
      "content": "<p>Thanks god, nearly no shake up. Happy new year!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1633635,
          "author_name": "theoviel",
          "author_url": "",
          "post_date": "12/31/2021 00:36:26",
          "content": "<p>🙁                </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1633717,
          "author_name": "steamedsheep",
          "author_url": "",
          "post_date": "12/31/2021 02:08:12",
          "content": "<p>Sorry for your shake down, good luck next time!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1633615,
      "author_name": "woprime",
      "author_url": "",
      "post_date": "12/31/2021 00:21:14",
      "content": "<p>Grats to all the winners, looking forward to your solutions! </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1631958": "Will there be a big shake up or just little, what do you think?",
    "1631961": "The private testset is small. \nquite big shake up for some teams, but not the top ones I think.",
    "1631979": "If one overfit the public leaderboard, it will be much easier for him to reach higher LB. So I think the chance of shaking up is even.",
    "1632008": "I think a shake of about 0.005 is possible, depending on \"cort\" prevalence.",
    "1632028": "±5-10 in gold zone, ±10-20 silver zone, ±50 in bronze zone",
    "1632039": "I hope no shake.",
    "1632054": "This is exactly what I hope too.",
    "1632120": "I think @takuok and @tascj0 will stay 1st if they keep their 0.003 lead on the public LB, as I expect a +/- 0.002 variation. \nHowever 0.345 is not that far from 0.343, the rest of the top 5 could change a bit !\nMy assumption is that organizers did a good job splitting the data :)\n\nOverall the private scores may be lower than public scores as I believe cort cells are over-represented in the public LB (which explains LB > CV).",
    "1632230": "I hope no shakeup with my team :))",
    "1632262": "I pray for a shakeup > 0.012~><~",
    "1632523": "Hoping for a shake up for us =)",
    "1632592": "Cant agree more",
    "1632645": "If there is a shake up, it will most probably be caused by a shift in class prevalence. This could be the case like @theoviel mentionned because cort was clearly over-represented in public LB compared to training distribution.",
    "1632943": "wait ,first tell me what is this Shake :>",
    "1633327": "It depends on the distribution of the three classes(shsy5y, cort, astro), in the public LB cort may have more which contribute to higher score.And shsy5y and astro is easily being overfitting.Hope a shake up to a higher rank🙏",
    "1633443": "balavashan Shakeup means a change in rankings when private ranking is calculated",
    "1633607": "Thanks god, nearly no shake up. Happy new year!",
    "1633615": "Grats to all the winners, looking forward to your solutions!",
    "1633635": "🙁",
    "1633717": "Sorry for your shake down, good luck next time!"
  },
  "source": "meta"
}