{
  "id": 550492,
  "title": "Test data order",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/550492",
  "author_name": "Big_Hammer_zz",
  "post_date": "2024-12-08T00:45:47.584000",
  "votes": 0,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I was trying to use a moving average and some historical lag data to improve my prediction. Is it guaranteed that the test data immediately follows the training data in time order? Specifically, can I assume that the test set starts the day after the last date_id in the training set?</p>\n<p>If not, how should I handle the gap between the training and test data when incorporating lag features or moving averages into my model?</p>",
  "messages": [
    {
      "id": 3066294,
      "postDate": "2024-12-08T00:57:47.747Z",
      "content": "<p>I think this has been discussed previously and the answer was yes the test data follows on chronologically from the train because the lags of the responders of the first day in the test matched the responder values in the last day of the train data</p>",
      "rawMarkdown": "I think this has been discussed previously and the answer was yes the test data follows on chronologically from the train because the lags of the responders of the first day in the test matched the responder values in the last day of the train data",
      "votes": 1,
      "replies": [
        {
          "id": 3066662,
          "postDate": "2024-12-08T10:36:01.820Z",
          "content": "<p>Thank you because I can't find the related discussion at the first time. </p>",
          "rawMarkdown": "Thank you because I can't find the related discussion at the first time. "
        },
        {
          "id": 3066869,
          "postDate": "2024-12-08T14:54:06.193Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 3067206,
      "postDate": "2024-12-09T03:27:23.363Z",
      "content": "<p>I'm wondering the same thing…</p>\n<p>\"During the forecasting phase, the evaluation API will serve test data from the beginning of the public set to the end of the private set. You must make predictions at every timestep, but, in this phase, only predictions on the private set are scored. (You may predict 0.0 on the unscored segments, if you like.)\"</p>\n<ul>\n<li>makes it sound like it will… not sure if the public set is the training, or test, if it's the test that's only 1 time step</li>\n</ul>\n<p>\"Starting after the final submission deadline there will be periodic updates to the leaderboard to reflect market data updates that will be run against selected notebooks.\"</p>\n<ul>\n<li>makes it sound like it won't…</li>\n</ul>",
      "rawMarkdown": "I'm wondering the same thing...\n\n\"During the forecasting phase, the evaluation API will serve test data from the beginning of the public set to the end of the private set. You must make predictions at every timestep, but, in this phase, only predictions on the private set are scored. (You may predict 0.0 on the unscored segments, if you like.)\"\n\n- makes it sound like it will... not sure if the public set is the training, or test, if it's the test that's only 1 time step\n\n\"Starting after the final submission deadline there will be periodic updates to the leaderboard to reflect market data updates that will be run against selected notebooks.\"\n\n- makes it sound like it won't...",
      "votes": 2,
      "replies": [
        {
          "id": 3068126,
          "postDate": "2024-12-10T01:46:33.003Z",
          "content": "<p>I just check the lags from api  and it match the value in trainset date_id = 1698 and time_id = 0</p>",
          "rawMarkdown": "I just check the lags from api  and it match the value in trainset date_id = 1698 and time_id = 0\n",
          "votes": 1
        }
      ]
    },
    {
      "id": 3066285,
      "postDate": "2024-12-08T00:45:47.583Z",
      "content": "<p>I was trying to use a moving average and some historical lag data to improve my prediction. Is it guaranteed that the test data immediately follows the training data in time order? Specifically, can I assume that the test set starts the day after the last date_id in the training set?</p>\n<p>If not, how should I handle the gap between the training and test data when incorporating lag features or moving averages into my model?</p>",
      "rawMarkdown": "I was trying to use a moving average and some historical lag data to improve my prediction. Is it guaranteed that the test data immediately follows the training data in time order? Specifically, can I assume that the test set starts the day after the last date_id in the training set?\n\nIf not, how should I handle the gap between the training and test data when incorporating lag features or moving averages into my model?"
    }
  ],
  "comments": [
    {
      "id": 3066294,
      "author_name": "LGreig",
      "author_url": "",
      "post_date": "2024-12-08T00:57:47.747000",
      "content": "<p>I think this has been discussed previously and the answer was yes the test data follows on chronologically from the train because the lags of the responders of the first day in the test matched the responder values in the last day of the train data</p>",
      "votes": 1,
      "replies": [
        {
          "id": 3066662,
          "author_name": "Big_Hammer_zz",
          "author_url": "",
          "post_date": "2024-12-08T10:36:01.820000",
          "content": "<p>Thank you because I can't find the related discussion at the first time. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 3066869,
          "author_name": "",
          "author_url": "",
          "post_date": "2024-12-08T14:54:06.193000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 3067206,
      "author_name": "winddude",
      "author_url": "",
      "post_date": "2024-12-09T03:27:23.363000",
      "content": "<p>I'm wondering the same thing…</p>\n<p>\"During the forecasting phase, the evaluation API will serve test data from the beginning of the public set to the end of the private set. You must make predictions at every timestep, but, in this phase, only predictions on the private set are scored. (You may predict 0.0 on the unscored segments, if you like.)\"</p>\n<ul>\n<li>makes it sound like it will… not sure if the public set is the training, or test, if it's the test that's only 1 time step</li>\n</ul>\n<p>\"Starting after the final submission deadline there will be periodic updates to the leaderboard to reflect market data updates that will be run against selected notebooks.\"</p>\n<ul>\n<li>makes it sound like it won't…</li>\n</ul>",
      "votes": 2,
      "replies": [
        {
          "id": 3068126,
          "author_name": "Big_Hammer_zz",
          "author_url": "",
          "post_date": "2024-12-10T01:46:33.003000",
          "content": "<p>I just check the lags from api  and it match the value in trainset date_id = 1698 and time_id = 0</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3066294": "I think this has been discussed previously and the answer was yes the test data follows on chronologically from the train because the lags of the responders of the first day in the test matched the responder values in the last day of the train data",
    "3067206": "I'm wondering the same thing...\n\n\"During the forecasting phase, the evaluation API will serve test data from the beginning of the public set to the end of the private set. You must make predictions at every timestep, but, in this phase, only predictions on the private set are scored. (You may predict 0.0 on the unscored segments, if you like.)\"\n\n- makes it sound like it will... not sure if the public set is the training, or test, if it's the test that's only 1 time step\n\n\"Starting after the final submission deadline there will be periodic updates to the leaderboard to reflect market data updates that will be run against selected notebooks.\"\n\n- makes it sound like it won't...",
    "3066285": "I was trying to use a moving average and some historical lag data to improve my prediction. Is it guaranteed that the test data immediately follows the training data in time order? Specifically, can I assume that the test set starts the day after the last date_id in the training set?\n\nIf not, how should I handle the gap between the training and test data when incorporating lag features or moving averages into my model?"
  }
}