{
  "id": 303747,
  "title": "a nuisance to be cleared off,, or is it already handled?",
  "url": "/competitions/g-research-crypto-forecasting/discussion/303747",
  "author_name": "",
  "post_date": "2022-01-29T05:56:31.389489Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Host says: 'we will use the same data processing during the evaluation phase'. <br>\nWhat does this mean for 'log return calculation'? </p>\n<ol>\n<li><p>Host will comply with the <strong>initial target creation process</strong> and utilize (x-1) approximation. (Here log(x) becomes x-1 for x very close to 1 (and this comes from Taylor expansion as explained in earlier posts), which appears to be used for computing efficiency.)</p></li>\n<li><p>Host will comply with the <strong>'competition documents'</strong> and use np.log(x) (or equivalent) for all log calculations.</p></li>\n</ol>",
  "messages": [
    {
      "id": "1667947",
      "postDate": "01/29/2022 05:56:31",
      "content": "<p>Host says: 'we will use the same data processing during the evaluation phase'. <br>\nWhat does this mean for 'log return calculation'? </p>\n<ol>\n<li><p>Host will comply with the <strong>initial target creation process</strong> and utilize (x-1) approximation. (Here log(x) becomes x-1 for x very close to 1 (and this comes from Taylor expansion as explained in earlier posts), which appears to be used for computing efficiency.)</p></li>\n<li><p>Host will comply with the <strong>'competition documents'</strong> and use np.log(x) (or equivalent) for all log calculations.</p></li>\n</ol>",
      "rawMarkdown": "Host says: 'we will use the same data processing during the evaluation phase'. \nWhat does this mean for 'log return calculation'? \n\n1. Host will comply with the **initial target creation process** and utilize (x-1) approximation. (Here log(x) becomes x-1 for x very close to 1 (and this comes from Taylor expansion as explained in earlier posts), which appears to be used for computing efficiency.)\n\n2. Host will comply with the **'competition documents'** and use np.log(x) (or equivalent) for all log calculations.",
      "votes": null
    },
    {
      "id": "1668322",
      "postDate": "01/29/2022 14:21:07",
      "content": "<p>The additional supplemental_train Targets are consistent with option 1 - that is the version I am sticking with. Option 1 is also robust to the (I admit, odd) situation where the Close price goes negative.</p>",
      "rawMarkdown": "The additional supplemental_train Targets are consistent with option 1 - that is the version I am sticking with. Option 1 is also robust to the (I admit, odd) situation where the Close price goes negative.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1668322,
      "author_name": "jagofc",
      "author_url": "",
      "post_date": "01/29/2022 14:21:07",
      "content": "<p>The additional supplemental_train Targets are consistent with option 1 - that is the version I am sticking with. Option 1 is also robust to the (I admit, odd) situation where the Close price goes negative.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1667947": "Host says: 'we will use the same data processing during the evaluation phase'. \nWhat does this mean for 'log return calculation'? \n\n1. Host will comply with the **initial target creation process** and utilize (x-1) approximation. (Here log(x) becomes x-1 for x very close to 1 (and this comes from Taylor expansion as explained in earlier posts), which appears to be used for computing efficiency.)\n\n2. Host will comply with the **'competition documents'** and use np.log(x) (or equivalent) for all log calculations.",
    "1668322": "The additional supplemental_train Targets are consistent with option 1 - that is the version I am sticking with. Option 1 is also robust to the (I admit, odd) situation where the Close price goes negative."
  },
  "source": "meta"
}