{
  "id": 505620,
  "title": "The skyrocketing public  scores highlight just how bad of a competition this was",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/505620",
  "author_name": "ZhangYuX1012",
  "post_date": "2024-05-18T09:21:45.585000",
  "votes": 10,
  "comment_count": 2,
  "views": 0,
  "content": "<p>This competitions is so ridiculous that the highest score of public notebook went from 0.59 to 0. 645 in just one day, just by a few lines of \"metric hack\"🤣<br>\nMany people don't even have to  code to get themselves a silver medal with a simple click of \"edit\" and \"submit\"😄</p>",
  "messages": [
    {
      "id": 2821835,
      "postDate": "2024-05-18T09:21:45.587Z",
      "content": "<p>This competitions is so ridiculous that the highest score of public notebook went from 0.59 to 0. 645 in just one day, just by a few lines of \"metric hack\"🤣<br>\nMany people don't even have to  code to get themselves a silver medal with a simple click of \"edit\" and \"submit\"😄</p>",
      "rawMarkdown": "This competitions is so ridiculous that the highest score of public notebook went from 0.59 to 0. 645 in just one day, just by a few lines of \"metric hack\"🤣\nMany people don't even have to  code to get themselves a silver medal with a simple click of \"edit\" and \"submit\"😄",
      "votes": 10
    },
    {
      "id": 2821854,
      "postDate": "2024-05-18T09:37:39.690Z",
      "content": "<p>I strongly believe that the hacks only work on the public leaderboard.</p>\n<p>I speculate that the public test data are the during-Covid data, where client behavior was the most \"unexpected\". Perhaps the train data and private test data represent pre-Covid and after-Covid data respectively, where client behavior is similar. Perhaps that's the reason the competition hosts and Kaggle did not take action on the recent metric hacks. </p>\n<p>(Note that I am neither trying to probe the test set distribution nor trying to end up like the Boeing whistleblowers - my speculation is nothing but a logical assumption based on the insights from other participants)</p>",
      "rawMarkdown": "I strongly believe that the hacks only work on the public leaderboard.\n\nI speculate that the public test data are the during-Covid data, where client behavior was the most \"unexpected\". Perhaps the train data and private test data represent pre-Covid and after-Covid data respectively, where client behavior is similar. Perhaps that's the reason the competition hosts and Kaggle did not take action on the recent metric hacks. \n\n(Note that I am neither trying to probe the test set distribution nor trying to end up like the Boeing whistleblowers - my speculation is nothing but a logical assumption based on the insights from other participants)",
      "votes": 7
    },
    {
      "id": 2821987,
      "postDate": "2024-05-18T11:06:46.167Z",
      "rawMarkdown": "",
      "votes": -1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2821854,
      "author_name": "Andreas Bisiadis",
      "author_url": "",
      "post_date": "2024-05-18T09:37:39.690000",
      "content": "<p>I strongly believe that the hacks only work on the public leaderboard.</p>\n<p>I speculate that the public test data are the during-Covid data, where client behavior was the most \"unexpected\". Perhaps the train data and private test data represent pre-Covid and after-Covid data respectively, where client behavior is similar. Perhaps that's the reason the competition hosts and Kaggle did not take action on the recent metric hacks. </p>\n<p>(Note that I am neither trying to probe the test set distribution nor trying to end up like the Boeing whistleblowers - my speculation is nothing but a logical assumption based on the insights from other participants)</p>",
      "votes": 7,
      "replies": []
    },
    {
      "id": 2821987,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-05-18T11:06:46.167000",
      "content": "",
      "votes": -1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2821835": "This competitions is so ridiculous that the highest score of public notebook went from 0.59 to 0. 645 in just one day, just by a few lines of \"metric hack\"🤣\nMany people don't even have to  code to get themselves a silver medal with a simple click of \"edit\" and \"submit\"😄",
    "2821854": "I strongly believe that the hacks only work on the public leaderboard.\n\nI speculate that the public test data are the during-Covid data, where client behavior was the most \"unexpected\". Perhaps the train data and private test data represent pre-Covid and after-Covid data respectively, where client behavior is similar. Perhaps that's the reason the competition hosts and Kaggle did not take action on the recent metric hacks. \n\n(Note that I am neither trying to probe the test set distribution nor trying to end up like the Boeing whistleblowers - my speculation is nothing but a logical assumption based on the insights from other participants)",
    "2821987": ""
  }
}