{
  "id": 555914,
  "title": "Does the test dataset date id always start from 0?",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/555914",
  "author_name": "madmax0404",
  "post_date": "2025-01-10T04:37:27.782000",
  "votes": 0,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Does it always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?</p>\n<p>For example, say date id 0 in the train dataset is 1/1/2000. Does the test dataset's date id start from 0 even if the date is 2/3/2015?</p>\n<p>Or, does it align with the train dataset like, if date id 0 in the train dataset is 1/1/2000, then date id 0 in the test dataset is also 1/1/2000?</p>\n<p>Thanks in advance.</p>",
  "messages": [
    {
      "id": 3093002,
      "postDate": "2025-01-10T11:03:07.150Z",
      "content": "<blockquote>\n  <p>Does [test dataset date id] always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?</p>\n</blockquote>\n<p>The data doc is vague about this, and I never trust deducing things from looking at sample .csv files if not explicitly said in the doc.</p>\n<p>The safest thing to do is to add C to all test date Id's, where C is a constant guaranteed to be larger than the largest training date Id.  At worst, we lose out on one test day's worth of online learning, but I doubt even this worst case applies.</p>",
      "rawMarkdown": ">Does [test dataset date id] always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?\n\nThe data doc is vague about this, and I never trust deducing things from looking at sample .csv files if not explicitly said in the doc.\n\nThe safest thing to do is to add C to all test date Id's, where C is a constant guaranteed to be larger than the largest training date Id.  At worst, we lose out on one test day's worth of online learning, but I doubt even this worst case applies.",
      "votes": 1
    },
    {
      "id": 3092804,
      "postDate": "2025-01-10T04:37:27.783Z",
      "content": "<p>Does it always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?</p>\n<p>For example, say date id 0 in the train dataset is 1/1/2000. Does the test dataset's date id start from 0 even if the date is 2/3/2015?</p>\n<p>Or, does it align with the train dataset like, if date id 0 in the train dataset is 1/1/2000, then date id 0 in the test dataset is also 1/1/2000?</p>\n<p>Thanks in advance.</p>",
      "rawMarkdown": "Does it always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?\n\nFor example, say date id 0 in the train dataset is 1/1/2000. Does the test dataset's date id start from 0 even if the date is 2/3/2015?\n\nOr, does it align with the train dataset like, if date id 0 in the train dataset is 1/1/2000, then date id 0 in the test dataset is also 1/1/2000?\n\nThanks in advance."
    },
    {
      "id": 3093008,
      "postDate": "2025-01-10T11:12:49.127Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 3093002,
      "author_name": "Truth Seeker",
      "author_url": "",
      "post_date": "2025-01-10T11:03:07.150000",
      "content": "<blockquote>\n  <p>Does [test dataset date id] always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?</p>\n</blockquote>\n<p>The data doc is vague about this, and I never trust deducing things from looking at sample .csv files if not explicitly said in the doc.</p>\n<p>The safest thing to do is to add C to all test date Id's, where C is a constant guaranteed to be larger than the largest training date Id.  At worst, we lose out on one test day's worth of online learning, but I doubt even this worst case applies.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 3093008,
      "author_name": "",
      "author_url": "",
      "post_date": "2025-01-10T11:12:49.127000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3093002": ">Does [test dataset date id] always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?\n\nThe data doc is vague about this, and I never trust deducing things from looking at sample .csv files if not explicitly said in the doc.\n\nThe safest thing to do is to add C to all test date Id's, where C is a constant guaranteed to be larger than the largest training date Id.  At worst, we lose out on one test day's worth of online learning, but I doubt even this worst case applies.",
    "3092804": "Does it always start from 0, not related to the train dataset, or does it align with the date ids in the train dataset?\n\nFor example, say date id 0 in the train dataset is 1/1/2000. Does the test dataset's date id start from 0 even if the date is 2/3/2015?\n\nOr, does it align with the train dataset like, if date id 0 in the train dataset is 1/1/2000, then date id 0 in the test dataset is also 1/1/2000?\n\nThanks in advance.",
    "3093008": ""
  }
}