{
  "id": 553685,
  "title": "Clarity on format of prediction horizon",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/553685",
  "author_name": "",
  "post_date": "2024-12-27T19:30:49.862449800Z",
  "votes": -2,
  "comment_count": 2,
  "views": 0,
  "content": "<ol>\n<li>For predicting responder 6, is it expected this will be exactly one value for the entire day? I.e. we are making a forward prediction for the next date_id?</li>\n<li>What is row_id? Does this just map to the unique symbols? That is, each symbol has it's own responder_6 we are predicting?</li>\n<li>How far back are we allowed to use? Is it only the previous day's data, or is it essentially unlimited? For example, training on 15 days of data for a +1 day prediction.</li>\n</ol>",
  "messages": [
    {
      "id": "3082194",
      "postDate": "12/27/2024 19:30:49",
      "content": "<ol>\n<li>For predicting responder 6, is it expected this will be exactly one value for the entire day? I.e. we are making a forward prediction for the next date_id?</li>\n<li>What is row_id? Does this just map to the unique symbols? That is, each symbol has it's own responder_6 we are predicting?</li>\n<li>How far back are we allowed to use? Is it only the previous day's data, or is it essentially unlimited? For example, training on 15 days of data for a +1 day prediction.</li>\n</ol>",
      "rawMarkdown": "1. For predicting responder 6, is it expected this will be exactly one value for the entire day? I.e. we are making a forward prediction for the next date_id?\n2. What is row_id? Does this just map to the unique symbols? That is, each symbol has it's own responder_6 we are predicting?\n3. How far back are we allowed to use? Is it only the previous day's data, or is it essentially unlimited? For example, training on 15 days of data for a +1 day prediction.",
      "votes": null
    },
    {
      "id": "3082375",
      "postDate": "12/28/2024 04:27:48",
      "content": "<p>We can use any length of the lagged data so long as you have the requisite data <a href=\"https://www.kaggle.com/xraygoth\" target=\"_blank\">@xraygoth</a> <br>\nRow_id is just an identifier for the row and has no other significance. <br>\nResponder_6 will have different values for different symbols and time_id and is not constant </p>",
      "rawMarkdown": "We can use any length of the lagged data so long as you have the requisite data @xraygoth \nRow_id is just an identifier for the row and has no other significance. \nResponder_6 will have different values for different symbols and time_id and is not constant",
      "votes": null
    },
    {
      "id": "3083544",
      "postDate": "12/29/2024 16:20:49",
      "content": "<p>I appreciate the comment. One thing that is still unclear to me is how far forward each prediction is supposed to encompass. Are we predicting the value of responder_6 at the t+1 time_id, which is quite granular, or are we predicting responder_6 for the entire t+1 date_id? Also, about the row_id comment. There are 39 rows, so each row should correspond to the relevant symbol as each one as its own responder_6, correct?</p>",
      "rawMarkdown": "I appreciate the comment. One thing that is still unclear to me is how far forward each prediction is supposed to encompass. Are we predicting the value of responder_6 at the t+1 time_id, which is quite granular, or are we predicting responder_6 for the entire t+1 date_id? Also, about the row_id comment. There are 39 rows, so each row should correspond to the relevant symbol as each one as its own responder_6, correct?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3082375,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "12/28/2024 04:27:48",
      "content": "<p>We can use any length of the lagged data so long as you have the requisite data <a href=\"https://www.kaggle.com/xraygoth\" target=\"_blank\">@xraygoth</a> <br>\nRow_id is just an identifier for the row and has no other significance. <br>\nResponder_6 will have different values for different symbols and time_id and is not constant </p>",
      "votes": null,
      "replies": [
        {
          "id": 3083544,
          "author_name": "xraygoth",
          "author_url": "",
          "post_date": "12/29/2024 16:20:49",
          "content": "<p>I appreciate the comment. One thing that is still unclear to me is how far forward each prediction is supposed to encompass. Are we predicting the value of responder_6 at the t+1 time_id, which is quite granular, or are we predicting responder_6 for the entire t+1 date_id? Also, about the row_id comment. There are 39 rows, so each row should correspond to the relevant symbol as each one as its own responder_6, correct?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3082194": "1. For predicting responder 6, is it expected this will be exactly one value for the entire day? I.e. we are making a forward prediction for the next date_id?\n2. What is row_id? Does this just map to the unique symbols? That is, each symbol has it's own responder_6 we are predicting?\n3. How far back are we allowed to use? Is it only the previous day's data, or is it essentially unlimited? For example, training on 15 days of data for a +1 day prediction.",
    "3082375": "We can use any length of the lagged data so long as you have the requisite data @xraygoth \nRow_id is just an identifier for the row and has no other significance. \nResponder_6 will have different values for different symbols and time_id and is not constant",
    "3083544": "I appreciate the comment. One thing that is still unclear to me is how far forward each prediction is supposed to encompass. Are we predicting the value of responder_6 at the t+1 time_id, which is quite granular, or are we predicting responder_6 for the entire t+1 date_id? Also, about the row_id comment. There are 39 rows, so each row should correspond to the relevant symbol as each one as its own responder_6, correct?"
  },
  "source": "meta"
}