{
  "id": 112896,
  "title": "How are the public LB and private LB spitted? ",
  "url": "/competitions/ashrae-energy-prediction/discussion/112896",
  "author_name": "",
  "post_date": "2019-10-16T02:44:36.536840900Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Are they splitted by time? Or just randomly pick 78% testing samples as the public LB?</p>",
  "messages": [
    {
      "id": "650022",
      "postDate": "10/16/2019 02:44:36",
      "content": "<p>Are they splitted by time? Or just randomly pick 78% testing samples as the public LB?</p>",
      "rawMarkdown": "Are they splitted by time? Or just randomly pick 78% testing samples as the public LB?",
      "votes": null
    },
    {
      "id": "650030",
      "postDate": "10/16/2019 02:56:00",
      "content": "<p>Usually, this information isn't given by the competition host and you should use some submissions to probe the leaderboard and try to find out!</p>",
      "rawMarkdown": "Usually, this information isn't given by the competition host and you should use some submissions to probe the leaderboard and try to find out!",
      "votes": null
    },
    {
      "id": "650033",
      "postDate": "10/16/2019 02:59:04",
      "content": "<p>That's true...\nBut I think they might randomly select 78% data as public LB since my validation score and LB score are very close</p>\n\n<h1>Edit:</h1>\n\n<p>I'm wrong. It's split by time </p>",
      "rawMarkdown": "That's true...\nBut I think they might randomly select 78% data as public LB since my validation score and LB score are very close\n\n# Edit:\nI'm wrong. It's split by time",
      "votes": null
    },
    {
      "id": "650035",
      "postDate": "10/16/2019 03:02:05",
      "content": "<p>Even if it is randomly split just probe the leaderboard :)</p>",
      "rawMarkdown": "Even if it is randomly split just probe the leaderboard :)",
      "votes": null
    },
    {
      "id": "650037",
      "postDate": "10/16/2019 03:03:53",
      "content": "<blockquote>\n  <p><strong>Inversion:</strong>\n  We always split time-based data by <strong>time</strong> for Train/Test, and Public/Private. If we didn't do that, the Public leaderboard score would provide feedback (leakage) for future time points.</p>\n</blockquote>",
      "rawMarkdown": "&gt; **Inversion:**\n&gt; We always split time-based data by **time** for Train/Test, and Public/Private. If we didn't do that, the Public leaderboard score would provide feedback (leakage) for future time points.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 650030,
      "author_name": "tayorm",
      "author_url": "",
      "post_date": "10/16/2019 02:56:00",
      "content": "<p>Usually, this information isn't given by the competition host and you should use some submissions to probe the leaderboard and try to find out!</p>",
      "votes": null,
      "replies": [
        {
          "id": 650033,
          "author_name": "alex495240115",
          "author_url": "",
          "post_date": "10/16/2019 02:59:04",
          "content": "<p>That's true...\nBut I think they might randomly select 78% data as public LB since my validation score and LB score are very close</p>\n\n<h1>Edit:</h1>\n\n<p>I'm wrong. It's split by time </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 650035,
          "author_name": "tayorm",
          "author_url": "",
          "post_date": "10/16/2019 03:02:05",
          "content": "<p>Even if it is randomly split just probe the leaderboard :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 650037,
      "author_name": "authman",
      "author_url": "",
      "post_date": "10/16/2019 03:03:53",
      "content": "<blockquote>\n  <p><strong>Inversion:</strong>\n  We always split time-based data by <strong>time</strong> for Train/Test, and Public/Private. If we didn't do that, the Public leaderboard score would provide feedback (leakage) for future time points.</p>\n</blockquote>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "650022": "Are they splitted by time? Or just randomly pick 78% testing samples as the public LB?",
    "650030": "Usually, this information isn't given by the competition host and you should use some submissions to probe the leaderboard and try to find out!",
    "650033": "That's true...\nBut I think they might randomly select 78% data as public LB since my validation score and LB score are very close\n\n# Edit:\nI'm wrong. It's split by time",
    "650035": "Even if it is randomly split just probe the leaderboard :)",
    "650037": "&gt; **Inversion:**\n&gt; We always split time-based data by **time** for Train/Test, and Public/Private. If we didn't do that, the Public leaderboard score would provide feedback (leakage) for future time points."
  },
  "source": "meta"
}