{
  "id": 189272,
  "title": "why can't I choose the notebook that performs best  on the private dataset ?",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/189272",
  "author_name": "",
  "post_date": "2020-10-07T06:09:42.815638500Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I can tell that the notebook chosen for the public score and private score is the same, however, the notebook automatically chosen is the one that performed best on the public dataset, I have other notebooks that performed so well on the private dataset but not as well on the public one. I am unable to select these notebooks, when I try to do so , it says \"manual selection is currently disabled\", why is that?<br>\nwhy can't I choose the notebook that performs best on the private dataset rather than one that performs best on the public one, will the selection be enabled soon?</p>",
  "messages": [
    {
      "id": "1040418",
      "postDate": "10/07/2020 06:09:42",
      "content": "<p>I can tell that the notebook chosen for the public score and private score is the same, however, the notebook automatically chosen is the one that performed best on the public dataset, I have other notebooks that performed so well on the private dataset but not as well on the public one. I am unable to select these notebooks, when I try to do so , it says \"manual selection is currently disabled\", why is that?<br>\nwhy can't I choose the notebook that performs best on the private dataset rather than one that performs best on the public one, will the selection be enabled soon?</p>",
      "rawMarkdown": "I can tell that the notebook chosen for the public score and private score is the same, however, the notebook automatically chosen is the one that performed best on the public dataset, I have other notebooks that performed so well on the private dataset but not as well on the public one. I am unable to select these notebooks, when I try to do so , it says \"manual selection is currently disabled\", why is that?\nwhy can't I choose the notebook that performs best on the private dataset rather than one that performs best on the public one, will the selection be enabled soon?",
      "votes": null
    },
    {
      "id": "1040437",
      "postDate": "10/07/2020 06:19:42",
      "content": "<p>You can only select your submissions manually before the release of private leaderboard. If you won't choose any submissions, kaggle would automatically submit your best 2 public scoring submissions in the private leaderboard. Since competition is over now, this is going to be your final rank.(maybe little bit change in rank after removing cheaters and cleaning leaderbaord)</p>",
      "rawMarkdown": "You can only select your submissions manually before the release of private leaderboard. If you won't choose any submissions, kaggle would automatically submit your best 2 public scoring submissions in the private leaderboard. Since competition is over now, this is going to be your final rank.(maybe little bit change in rank after removing cheaters and cleaning leaderbaord)",
      "votes": null
    },
    {
      "id": "1040482",
      "postDate": "10/07/2020 06:55:07",
      "content": "<p>That'd give you an advantage for constantly submitting, and would pretty much be overfitting to the private dataset. If you submitted 500 times with random hyperparameters one of them might get a really good score but wouldn't actually be useful.</p>",
      "rawMarkdown": "That'd give you an advantage for constantly submitting, and would pretty much be overfitting to the private dataset. If you submitted 500 times with random hyperparameters one of them might get a really good score but wouldn't actually be useful.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1040437,
      "author_name": "nischaydnk",
      "author_url": "",
      "post_date": "10/07/2020 06:19:42",
      "content": "<p>You can only select your submissions manually before the release of private leaderboard. If you won't choose any submissions, kaggle would automatically submit your best 2 public scoring submissions in the private leaderboard. Since competition is over now, this is going to be your final rank.(maybe little bit change in rank after removing cheaters and cleaning leaderbaord)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1040482,
      "author_name": "humphreymunn",
      "author_url": "",
      "post_date": "10/07/2020 06:55:07",
      "content": "<p>That'd give you an advantage for constantly submitting, and would pretty much be overfitting to the private dataset. If you submitted 500 times with random hyperparameters one of them might get a really good score but wouldn't actually be useful.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1040418": "I can tell that the notebook chosen for the public score and private score is the same, however, the notebook automatically chosen is the one that performed best on the public dataset, I have other notebooks that performed so well on the private dataset but not as well on the public one. I am unable to select these notebooks, when I try to do so , it says \"manual selection is currently disabled\", why is that?\nwhy can't I choose the notebook that performs best on the private dataset rather than one that performs best on the public one, will the selection be enabled soon?",
    "1040437": "You can only select your submissions manually before the release of private leaderboard. If you won't choose any submissions, kaggle would automatically submit your best 2 public scoring submissions in the private leaderboard. Since competition is over now, this is going to be your final rank.(maybe little bit change in rank after removing cheaters and cleaning leaderbaord)",
    "1040482": "That'd give you an advantage for constantly submitting, and would pretty much be overfitting to the private dataset. If you submitted 500 times with random hyperparameters one of them might get a really good score but wouldn't actually be useful."
  },
  "source": "meta"
}