{
  "id": 587792,
  "title": "Possibility to train on paths",
  "url": "/competitions/drw-crypto-market-prediction/discussion/587792",
  "author_name": "Lii_Maxime",
  "post_date": "2025-07-02T18:35:26.247000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Usually models train on historical data for a given period. I think that this can be implemented here for the train, but useless for prediction as the time stamp in the test are removed. Any idea for these type of model to harness more classical methods?</p>",
  "messages": [
    {
      "id": 3239368,
      "postDate": "2025-07-02T18:35:26.247Z",
      "content": "<p>Usually models train on historical data for a given period. I think that this can be implemented here for the train, but useless for prediction as the time stamp in the test are removed. Any idea for these type of model to harness more classical methods?</p>",
      "rawMarkdown": "Usually models train on historical data for a given period. I think that this can be implemented here for the train, but useless for prediction as the time stamp in the test are removed. Any idea for these type of model to harness more classical methods?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3239368": "Usually models train on historical data for a given period. I think that this can be implemented here for the train, but useless for prediction as the time stamp in the test are removed. Any idea for these type of model to harness more classical methods?"
  }
}