{
  "id": 506061,
  "title": "I got a score less than 0。",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/506061",
  "author_name": "",
  "post_date": "2024-05-20T10:05:20.455919100Z",
  "votes": 3,
  "comment_count": 5,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15707707%2F39fc2eedb0b73953dcb34217c1fa25f3%2Fba75c32ce47c9644353bc986ebfdd26.png?generation=1716199518066979&amp;alt=media\"></p>",
  "messages": [
    {
      "id": "2825327",
      "postDate": "05/20/2024 10:05:20",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15707707%2F39fc2eedb0b73953dcb34217c1fa25f3%2Fba75c32ce47c9644353bc986ebfdd26.png?generation=1716199518066979&amp;alt=media\"></p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15707707%2F39fc2eedb0b73953dcb34217c1fa25f3%2Fba75c32ce47c9644353bc986ebfdd26.png?generation=1716199518066979&alt=media)",
      "votes": null
    },
    {
      "id": "2825338",
      "postDate": "05/20/2024 10:12:45",
      "content": "<p>Based on a solution with a score of 0.657, but when the metric was slightly altered, it was penalized with negative points.</p>",
      "rawMarkdown": "Based on a solution with a score of 0.657, but when the metric was slightly altered, it was penalized with negative points.",
      "votes": null
    },
    {
      "id": "2826149",
      "postDate": "05/20/2024 18:19:48",
      "content": "<p>Clear evidence that metric tricks can be painful on private lb. I recommend doing a simple ensemble of boustings and not adjusting the result by metrics, as this is likely to be fraught with severe overfitting.</p>",
      "rawMarkdown": "Clear evidence that metric tricks can be painful on private lb. I recommend doing a simple ensemble of boustings and not adjusting the result by metrics, as this is likely to be fraught with severe overfitting.",
      "votes": null
    },
    {
      "id": "2826518",
      "postDate": "05/21/2024 02:12:40",
      "content": "<p>Yes, metric tricks are sensitive, and small changes can make all the difference.</p>",
      "rawMarkdown": "Yes, metric tricks are sensitive, and small changes can make all the difference.",
      "votes": null
    },
    {
      "id": "2828280",
      "postDate": "05/22/2024 02:38:57",
      "content": "<p>I cannot make my notebook public at the moment but I also have one with a negative score, and that was without metric hacking. What does a negative score mean mathematically?</p>",
      "rawMarkdown": "I cannot make my notebook public at the moment but I also have one with a negative score, and that was without metric hacking. What does a negative score mean mathematically?",
      "votes": null
    },
    {
      "id": "2829092",
      "postDate": "05/22/2024 12:15:40",
      "content": "<p>The stability penalty is too severe</p>",
      "rawMarkdown": "The stability penalty is too severe",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2825338,
      "author_name": "",
      "author_url": "",
      "post_date": "05/20/2024 10:12:45",
      "content": "<p>Based on a solution with a score of 0.657, but when the metric was slightly altered, it was penalized with negative points.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2826149,
          "author_name": "bratkovskyevgeny",
          "author_url": "",
          "post_date": "05/20/2024 18:19:48",
          "content": "<p>Clear evidence that metric tricks can be painful on private lb. I recommend doing a simple ensemble of boustings and not adjusting the result by metrics, as this is likely to be fraught with severe overfitting.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2826518,
              "author_name": "",
              "author_url": "",
              "post_date": "05/21/2024 02:12:40",
              "content": "<p>Yes, metric tricks are sensitive, and small changes can make all the difference.</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 2828280,
      "author_name": "rickpack",
      "author_url": "",
      "post_date": "05/22/2024 02:38:57",
      "content": "<p>I cannot make my notebook public at the moment but I also have one with a negative score, and that was without metric hacking. What does a negative score mean mathematically?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2829092,
          "author_name": "",
          "author_url": "",
          "post_date": "05/22/2024 12:15:40",
          "content": "<p>The stability penalty is too severe</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2825327": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F15707707%2F39fc2eedb0b73953dcb34217c1fa25f3%2Fba75c32ce47c9644353bc986ebfdd26.png?generation=1716199518066979&alt=media)",
    "2825338": "Based on a solution with a score of 0.657, but when the metric was slightly altered, it was penalized with negative points.",
    "2826149": "Clear evidence that metric tricks can be painful on private lb. I recommend doing a simple ensemble of boustings and not adjusting the result by metrics, as this is likely to be fraught with severe overfitting.",
    "2826518": "Yes, metric tricks are sensitive, and small changes can make all the difference.",
    "2828280": "I cannot make my notebook public at the moment but I also have one with a negative score, and that was without metric hacking. What does a negative score mean mathematically?",
    "2829092": "The stability penalty is too severe"
  },
  "source": "meta"
}