{
  "id": 589999,
  "title": "Test set and lagged features",
  "url": "/competitions/drw-crypto-market-prediction/discussion/589999",
  "author_name": "",
  "post_date": "2025-07-17T00:29:59.703968800Z",
  "votes": -1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello! This question might be very naive, but I could not find any answer from other posts. As stated by the organizers of the competition, the test set has been masked and shuffled. Doesn't that make it impossible to use lagged values of any kind? Therefore, our predictions must solely be based on all the other features we have at the time of the prediction? I have seen some codes where people were using lagged feature with various fortune, so I was wondering…</p>",
  "messages": [
    {
      "id": "3249699",
      "postDate": "07/17/2025 00:29:59",
      "content": "<p>Hello! This question might be very naive, but I could not find any answer from other posts. As stated by the organizers of the competition, the test set has been masked and shuffled. Doesn't that make it impossible to use lagged values of any kind? Therefore, our predictions must solely be based on all the other features we have at the time of the prediction? I have seen some codes where people were using lagged feature with various fortune, so I was wondering…</p>",
      "rawMarkdown": "Hello! This question might be very naive, but I could not find any answer from other posts. As stated by the organizers of the competition, the test set has been masked and shuffled. Doesn't that make it impossible to use lagged values of any kind? Therefore, our predictions must solely be based on all the other features we have at the time of the prediction? I have seen some codes where people were using lagged feature with various fortune, so I was wondering...",
      "votes": null
    },
    {
      "id": "3249878",
      "postDate": "07/17/2025 09:25:55",
      "content": "<p>You are correct. People have reverse engineered the shuffled timestamps and unshuffled the test data allowing them to used lagged features. I have no idea why they're doing this, it has been stated outright to very clearly be against the rules and will not count at final submission, still people are doing it.</p>",
      "rawMarkdown": "You are correct. People have reverse engineered the shuffled timestamps and unshuffled the test data allowing them to used lagged features. I have no idea why they're doing this, it has been stated outright to very clearly be against the rules and will not count at final submission, still people are doing it.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3249878,
      "author_name": "greatestcutie",
      "author_url": "",
      "post_date": "07/17/2025 09:25:55",
      "content": "<p>You are correct. People have reverse engineered the shuffled timestamps and unshuffled the test data allowing them to used lagged features. I have no idea why they're doing this, it has been stated outright to very clearly be against the rules and will not count at final submission, still people are doing it.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3249699": "Hello! This question might be very naive, but I could not find any answer from other posts. As stated by the organizers of the competition, the test set has been masked and shuffled. Doesn't that make it impossible to use lagged values of any kind? Therefore, our predictions must solely be based on all the other features we have at the time of the prediction? I have seen some codes where people were using lagged feature with various fortune, so I was wondering...",
    "3249878": "You are correct. People have reverse engineered the shuffled timestamps and unshuffled the test data allowing them to used lagged features. I have no idea why they're doing this, it has been stated outright to very clearly be against the rules and will not count at final submission, still people are doing it."
  },
  "source": "meta"
}