{
  "id": 297706,
  "title": "On-line target calculation",
  "url": "/competitions/g-research-crypto-forecasting/discussion/297706",
  "author_name": "",
  "post_date": "2021-12-29T00:29:08.530624600Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Did somebody already look into how to compute <code>Target</code> from <code>Close</code> in an on-line fashion using the API? There has been a vivid discussion on how to recreate <code>Target</code> of the <code>train.csv</code> but that approach/code will fail on the API. One point of concern is that Beta is computed as a rolling average over some time window. That won't be possible during inference time.</p>\n<p>Cheers!</p>",
  "messages": [
    {
      "id": "1631937",
      "postDate": "12/29/2021 00:29:08",
      "content": "<p>Did somebody already look into how to compute <code>Target</code> from <code>Close</code> in an on-line fashion using the API? There has been a vivid discussion on how to recreate <code>Target</code> of the <code>train.csv</code> but that approach/code will fail on the API. One point of concern is that Beta is computed as a rolling average over some time window. That won't be possible during inference time.</p>\n<p>Cheers!</p>",
      "rawMarkdown": "Did somebody already look into how to compute `Target` from `Close` in an on-line fashion using the API? There has been a vivid discussion on how to recreate `Target` of the `train.csv` but that approach/code will fail on the API. One point of concern is that Beta is computed as a rolling average over some time window. That won't be possible during inference time.\n\nCheers!",
      "votes": null
    },
    {
      "id": "1641533",
      "postDate": "01/07/2022 14:21:22",
      "content": "<p>Good question. Any progress?</p>",
      "rawMarkdown": "Good question. Any progress?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1641533,
      "author_name": "abdulkadirguner",
      "author_url": "",
      "post_date": "01/07/2022 14:21:22",
      "content": "<p>Good question. Any progress?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1631937": "Did somebody already look into how to compute `Target` from `Close` in an on-line fashion using the API? There has been a vivid discussion on how to recreate `Target` of the `train.csv` but that approach/code will fail on the API. One point of concern is that Beta is computed as a rolling average over some time window. That won't be possible during inference time.\n\nCheers!",
    "1641533": "Good question. Any progress?"
  },
  "source": "meta"
}