{
  "id": 290209,
  "title": "CV not correlated with lb without overlap ! Yet again",
  "url": "/competitions/g-research-crypto-forecasting/discussion/290209",
  "author_name": "Sayantan Mazumdar",
  "post_date": "2021-11-23T15:11:25.373000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>This is to inform the fact that even if we do not use the overlap we will still get CV not correlated with LB problems for newbies who dont know about it simply means that your public lb is not turning out to be similar to your score on the validation set this can happen for many reason namely two low validation set size which you are using or low number of samples in public lb , here i guess there 58 or something samples in public lb . Small enough there are 14 asset ids already . I would like to ask your public lb and CV scores cuz i spot a big difference </p>",
  "messages": [
    {
      "id": 1593029,
      "postDate": "2021-11-23T15:11:25.373Z",
      "content": "<p>Hi,</p>\n<p>This is to inform the fact that even if we do not use the overlap we will still get CV not correlated with LB problems for newbies who dont know about it simply means that your public lb is not turning out to be similar to your score on the validation set this can happen for many reason namely two low validation set size which you are using or low number of samples in public lb , here i guess there 58 or something samples in public lb . Small enough there are 14 asset ids already . I would like to ask your public lb and CV scores cuz i spot a big difference </p>",
      "rawMarkdown": "Hi,\n\nThis is to inform the fact that even if we do not use the overlap we will still get CV not correlated with LB problems for newbies who dont know about it simply means that your public lb is not turning out to be similar to your score on the validation set this can happen for many reason namely two low validation set size which you are using or low number of samples in public lb , here i guess there 58 or something samples in public lb . Small enough there are 14 asset ids already . I would like to ask your public lb and CV scores cuz i spot a big difference ",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1593029": "Hi,\n\nThis is to inform the fact that even if we do not use the overlap we will still get CV not correlated with LB problems for newbies who dont know about it simply means that your public lb is not turning out to be similar to your score on the validation set this can happen for many reason namely two low validation set size which you are using or low number of samples in public lb , here i guess there 58 or something samples in public lb . Small enough there are 14 asset ids already . I would like to ask your public lb and CV scores cuz i spot a big difference "
  }
}