{
  "id": 377894,
  "title": "Revealing the 4th digit of the score",
  "url": "/competitions/otto-recommender-system/discussion/377894",
  "author_name": "NikhilMishra",
  "post_date": "2023-01-13T10:35:58.396000",
  "votes": 13,
  "comment_count": 14,
  "views": 0,
  "content": "<p>Can kaggle reveal the 4th digit of the score for the competition ? With many people being stuck at the same score like 0.602, 0.6, 0.599,  I think it will be helpful to better track our progress. Otherwise the crowd is bound to grow even more <a href=\"https://www.kaggle.com/maggiemd\" target=\"_blank\">@maggiemd</a> </p>",
  "messages": [
    {
      "id": 2098223,
      "postDate": "2023-01-13T10:35:58.397Z",
      "content": "<p>Can kaggle reveal the 4th digit of the score for the competition ? With many people being stuck at the same score like 0.602, 0.6, 0.599,  I think it will be helpful to better track our progress. Otherwise the crowd is bound to grow even more <a href=\"https://www.kaggle.com/maggiemd\" target=\"_blank\">@maggiemd</a> </p>",
      "rawMarkdown": "Can kaggle reveal the 4th digit of the score for the competition ? With many people being stuck at the same score like 0.602, 0.6, 0.599,  I think it will be helpful to better track our progress. Otherwise the crowd is bound to grow even more @maggiemd ",
      "votes": 13
    },
    {
      "id": 2098465,
      "postDate": "2023-01-13T15:34:00.973Z",
      "content": "<p>In general, while the Public LB are useful, they are not always a good indicator of model accuracy. More decimal points are not done to avoid LB probing. Focus on improving your CV and choose the best CV. </p>",
      "rawMarkdown": "In general, while the Public LB are useful, they are not always a good indicator of model accuracy. More decimal points are not done to avoid LB probing. Focus on improving your CV and choose the best CV. ",
      "votes": 6,
      "replies": [
        {
          "id": 2098577,
          "postDate": "2023-01-13T17:06:59.400Z",
          "content": "<p>this time will take this advice .. recently i lost 4 sliver medal coz i ignore CV and choice the based on LB </p>",
          "rawMarkdown": "this time will take this advice .. recently i lost 4 sliver medal coz i ignore CV and choice the based on LB ",
          "votes": -1,
          "replies": [
            {
              "id": 2098897,
              "postDate": "2023-01-13T23:36:12.890Z",
              "content": "<p>Overfitting a leaderboard is much easier than finding a good solution</p>",
              "rawMarkdown": "Overfitting a leaderboard is much easier than finding a good solution"
            }
          ]
        }
      ]
    },
    {
      "id": 2106716,
      "postDate": "2023-01-19T10:18:13.120Z",
      "content": "<p>Poke <a href=\"https://www.kaggle.com/pnormann\" target=\"_blank\">@pnormann</a> ! <br>\nI think LB is reliable enough for us to have a 4th digit without worrying about people overfitting.</p>\n<p>It's gonna look quite awkward when the top 10 teams have the same 0.602 score otherwise :)</p>",
      "rawMarkdown": "Poke @pnormann ! \nI think LB is reliable enough for us to have a 4th digit without worrying about people overfitting.\n\nIt's gonna look quite awkward when the top 10 teams have the same 0.602 score otherwise :)",
      "votes": 1,
      "replies": [
        {
          "id": 2106759,
          "postDate": "2023-01-19T10:41:31.953Z",
          "content": "<p>or an or alternate solution, distribute prize money to all teams scoring 0.602 :D </p>",
          "rawMarkdown": "or an or alternate solution, distribute prize money to all teams scoring 0.602 :D "
        },
        {
          "id": 2107180,
          "postDate": "2023-01-19T16:09:10.093Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/theoviel\" target=\"_blank\">@theoviel</a>, we do not plan on increasing leaderboard precision until the competition has concluded. Competitors should rely on their local validation for model selection.</p>",
          "rawMarkdown": "Hi @theoviel, we do not plan on increasing leaderboard precision until the competition has concluded. Competitors should rely on their local validation for model selection.",
          "votes": 6,
          "replies": [
            {
              "id": 2107209,
              "postDate": "2023-01-19T16:25:07.017Z",
              "content": "<p>Makes sense!<br>\nThanks for the quick reply :)</p>",
              "rawMarkdown": "Makes sense!\nThanks for the quick reply :)",
              "votes": 1
            }
          ]
        }
      ]
    },
    {
      "id": 2104730,
      "postDate": "2023-01-18T02:05:20.530Z",
      "content": "<p>Can you take a loot into this? <a href=\"https://www.kaggle.com/maggiemd\" target=\"_blank\">@maggiemd</a> <a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> <a href=\"https://www.kaggle.com/sohier\" target=\"_blank\">@sohier</a> </p>",
      "rawMarkdown": "Can you take a loot into this? @maggiemd @inversion @sohier "
    },
    {
      "id": 2099046,
      "postDate": "2023-01-14T05:34:23.847Z",
      "content": "<p>I agree with you. The difference of 4th digit score is important for top teams.<br>\nRevealing the 4th digit of the score is helpful to develop worthwhile solutions.</p>",
      "rawMarkdown": "I agree with you. The difference of 4th digit score is important for top teams.\nRevealing the 4th digit of the score is helpful to develop worthwhile solutions."
    },
    {
      "id": 2098558,
      "postDate": "2023-01-13T16:49:07.843Z",
      "content": "<p>Agree, we have made many same-score submission but can only randomly guess their improvement.</p>",
      "rawMarkdown": "Agree, we have made many same-score submission but can only randomly guess their improvement.",
      "replies": [
        {
          "id": 2098570,
          "postDate": "2023-01-13T17:02:29.597Z",
          "content": "<p>exactly all I see is juggling of the top 3 spots without any improvement in the leaderboard :D</p>",
          "rawMarkdown": "exactly all I see is juggling of the top 3 spots without any improvement in the leaderboard :D"
        }
      ]
    },
    {
      "id": 2098255,
      "postDate": "2023-01-13T11:15:23.233Z",
      "content": "<p>I agree - that would be very helpful.</p>",
      "rawMarkdown": "I agree - that would be very helpful."
    },
    {
      "id": 2106763,
      "postDate": "2023-01-19T10:45:47.810Z",
      "rawMarkdown": "",
      "votes": -1,
      "isDeleted": true
    },
    {
      "id": 2099181,
      "postDate": "2023-01-14T08:41:00.210Z",
      "rawMarkdown": "",
      "votes": -2,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2098465,
      "author_name": "Kamal Das",
      "author_url": "",
      "post_date": "2023-01-13T15:34:00.973000",
      "content": "<p>In general, while the Public LB are useful, they are not always a good indicator of model accuracy. More decimal points are not done to avoid LB probing. Focus on improving your CV and choose the best CV. </p>",
      "votes": 6,
      "replies": [
        {
          "id": 2098577,
          "author_name": "AlmaamoonAl-saedi",
          "author_url": "",
          "post_date": "2023-01-13T17:06:59.400000",
          "content": "<p>this time will take this advice .. recently i lost 4 sliver medal coz i ignore CV and choice the based on LB </p>",
          "votes": -1,
          "replies": [
            {
              "id": 2098897,
              "author_name": "Thomas",
              "author_url": "",
              "post_date": "2023-01-13T23:36:12.890000",
              "content": "<p>Overfitting a leaderboard is much easier than finding a good solution</p>",
              "votes": 0,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2106716,
      "author_name": "Theo Viel",
      "author_url": "",
      "post_date": "2023-01-19T10:18:13.120000",
      "content": "<p>Poke <a href=\"https://www.kaggle.com/pnormann\" target=\"_blank\">@pnormann</a> ! <br>\nI think LB is reliable enough for us to have a 4th digit without worrying about people overfitting.</p>\n<p>It's gonna look quite awkward when the top 10 teams have the same 0.602 score otherwise :)</p>",
      "votes": 1,
      "replies": [
        {
          "id": 2106759,
          "author_name": "NikhilMishra",
          "author_url": "",
          "post_date": "2023-01-19T10:41:31.953000",
          "content": "<p>or an or alternate solution, distribute prize money to all teams scoring 0.602 :D </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 2107180,
          "author_name": "Philipp Normann",
          "author_url": "",
          "post_date": "2023-01-19T16:09:10.093000",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/theoviel\" target=\"_blank\">@theoviel</a>, we do not plan on increasing leaderboard precision until the competition has concluded. Competitors should rely on their local validation for model selection.</p>",
          "votes": 6,
          "replies": [
            {
              "id": 2107209,
              "author_name": "Theo Viel",
              "author_url": "",
              "post_date": "2023-01-19T16:25:07.017000",
              "content": "<p>Makes sense!<br>\nThanks for the quick reply :)</p>",
              "votes": 1,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2104730,
      "author_name": "ONODERA",
      "author_url": "",
      "post_date": "2023-01-18T02:05:20.530000",
      "content": "<p>Can you take a loot into this? <a href=\"https://www.kaggle.com/maggiemd\" target=\"_blank\">@maggiemd</a> <a href=\"https://www.kaggle.com/inversion\" target=\"_blank\">@inversion</a> <a href=\"https://www.kaggle.com/sohier\" target=\"_blank\">@sohier</a> </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2099046,
      "author_name": "toshi_k",
      "author_url": "",
      "post_date": "2023-01-14T05:34:23.847000",
      "content": "<p>I agree with you. The difference of 4th digit score is important for top teams.<br>\nRevealing the 4th digit of the score is helpful to develop worthwhile solutions.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2098558,
      "author_name": "Carno Zhao",
      "author_url": "",
      "post_date": "2023-01-13T16:49:07.843000",
      "content": "<p>Agree, we have made many same-score submission but can only randomly guess their improvement.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 2098570,
          "author_name": "NikhilMishra",
          "author_url": "",
          "post_date": "2023-01-13T17:02:29.597000",
          "content": "<p>exactly all I see is juggling of the top 3 spots without any improvement in the leaderboard :D</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 2098255,
      "author_name": "John Wakefield",
      "author_url": "",
      "post_date": "2023-01-13T11:15:23.233000",
      "content": "<p>I agree - that would be very helpful.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2106763,
      "author_name": "",
      "author_url": "",
      "post_date": "2023-01-19T10:45:47.810000",
      "content": "",
      "votes": -1,
      "replies": []
    },
    {
      "id": 2099181,
      "author_name": "",
      "author_url": "",
      "post_date": "2023-01-14T08:41:00.210000",
      "content": "",
      "votes": -2,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2098223": "Can kaggle reveal the 4th digit of the score for the competition ? With many people being stuck at the same score like 0.602, 0.6, 0.599,  I think it will be helpful to better track our progress. Otherwise the crowd is bound to grow even more @maggiemd ",
    "2098465": "In general, while the Public LB are useful, they are not always a good indicator of model accuracy. More decimal points are not done to avoid LB probing. Focus on improving your CV and choose the best CV. ",
    "2106716": "Poke @pnormann ! \nI think LB is reliable enough for us to have a 4th digit without worrying about people overfitting.\n\nIt's gonna look quite awkward when the top 10 teams have the same 0.602 score otherwise :)",
    "2104730": "Can you take a loot into this? @maggiemd @inversion @sohier ",
    "2099046": "I agree with you. The difference of 4th digit score is important for top teams.\nRevealing the 4th digit of the score is helpful to develop worthwhile solutions.",
    "2098558": "Agree, we have made many same-score submission but can only randomly guess their improvement.",
    "2098255": "I agree - that would be very helpful.",
    "2106763": "",
    "2099181": ""
  }
}