{
  "id": 47562,
  "title": "the reason for shaking up",
  "url": "/competitions/favorita-grocery-sales-forecasting/discussion/47562",
  "author_name": "",
  "post_date": "2018-01-16T10:25:39.526124300Z",
  "votes": 9,
  "comment_count": 2,
  "views": 0,
  "content": "<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/269155/8270/2018-01-16%2018-05-22%E5%B1%8F%E5%B9%95%E6%88%AA%E5%9B%BE.png\" alt=\"shake\"></p>\n\n<p>These two submissions was created by same model(ensembel of dozens lgbs with best local cv)</p>\n\n<p>The only difference between two submissions is whether to deal with the bias on promotion.</p>\n\n<p>the result:</p>\n\n<p>the teams do EDA seriously and try to handle the bias, drop in the leaderboard</p>\n\n<p>The teams just folk kernel and didn't realize the bias influence  ,jump in the leaderboard</p>\n\n<p>I'm very unhappy with this result.</p>\n\n<p><strong>I want to emphasize</strong> ：</p>\n\n<p>the reason for shake-up is not ignorance of local cv.</p>\n\n<p>I don't like that some people contribute their medal to the local cv</p>",
  "messages": [
    {
      "id": "269155",
      "postDate": "01/16/2018 10:25:39",
      "content": "<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/269155/8270/2018-01-16%2018-05-22%E5%B1%8F%E5%B9%95%E6%88%AA%E5%9B%BE.png\" alt=\"shake\"></p>\n\n<p>These two submissions was created by same model(ensembel of dozens lgbs with best local cv)</p>\n\n<p>The only difference between two submissions is whether to deal with the bias on promotion.</p>\n\n<p>the result:</p>\n\n<p>the teams do EDA seriously and try to handle the bias, drop in the leaderboard</p>\n\n<p>The teams just folk kernel and didn't realize the bias influence  ,jump in the leaderboard</p>\n\n<p>I'm very unhappy with this result.</p>\n\n<p><strong>I want to emphasize</strong> ：</p>\n\n<p>the reason for shake-up is not ignorance of local cv.</p>\n\n<p>I don't like that some people contribute their medal to the local cv</p>",
      "rawMarkdown": "![shake][1]\n\nThese two submissions was created by same model(ensembel of dozens lgbs with best local cv)\n\nThe only difference between two submissions is whether to deal with the bias on promotion.\n\nthe result:\n\n the teams do EDA seriously and try to handle the bias, drop in the leaderboard\n\nThe teams just folk kernel and didn't realize the bias influence  ,jump in the leaderboard\n\nI'm very unhappy with this result.\n\n**I want to emphasize** ：\n\nthe reason for shake-up is not ignorance of local cv.\n\n I don't like that some people contribute their medal to the local cv\n\n  [1]: https://kaggle2.blob.core.windows.net/forum-message-attachments/269155/8270/2018-01-16%2018-05-22%E5%B1%8F%E5%B9%95%E6%88%AA%E5%9B%BE.png",
      "votes": null
    },
    {
      "id": "269370",
      "postDate": "01/16/2018 17:45:22",
      "content": "<p>Cannot agree more with you. My best single model get 0.518 on pv and 0.510 on pb while my best average result get 0.515 on pv but 0.511 on pb. Similar local cv strategy just like other teams but not very helpful. Really a tragedy.</p>\n\n<p>（update: I make a big mistake on my cv set. So I am not sure whether there is a good cv set working for this competition.)</p>",
      "rawMarkdown": "Cannot agree more with you. My best single model get 0.518 on pv and 0.510 on pb while my best average result get 0.515 on pv but 0.511 on pb. Similar local cv strategy just like other teams but not very helpful. Really a tragedy.\n\n（update: I make a big mistake on my cv set. So I am not sure whether there is a good cv set working for this competition.)",
      "votes": null
    },
    {
      "id": "269386",
      "postDate": "01/16/2018 18:18:08",
      "content": "<p>Can you share how you tried to deal with bias?</p>",
      "rawMarkdown": "Can you share how you tried to deal with bias?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 269370,
      "author_name": "yedthon",
      "author_url": "",
      "post_date": "01/16/2018 17:45:22",
      "content": "<p>Cannot agree more with you. My best single model get 0.518 on pv and 0.510 on pb while my best average result get 0.515 on pv but 0.511 on pb. Similar local cv strategy just like other teams but not very helpful. Really a tragedy.</p>\n\n<p>（update: I make a big mistake on my cv set. So I am not sure whether there is a good cv set working for this competition.)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269386,
      "author_name": "npa02012",
      "author_url": "",
      "post_date": "01/16/2018 18:18:08",
      "content": "<p>Can you share how you tried to deal with bias?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "269155": "![shake][1]\n\nThese two submissions was created by same model(ensembel of dozens lgbs with best local cv)\n\nThe only difference between two submissions is whether to deal with the bias on promotion.\n\nthe result:\n\n the teams do EDA seriously and try to handle the bias, drop in the leaderboard\n\nThe teams just folk kernel and didn't realize the bias influence  ,jump in the leaderboard\n\nI'm very unhappy with this result.\n\n**I want to emphasize** ：\n\nthe reason for shake-up is not ignorance of local cv.\n\n I don't like that some people contribute their medal to the local cv\n\n  [1]: https://kaggle2.blob.core.windows.net/forum-message-attachments/269155/8270/2018-01-16%2018-05-22%E5%B1%8F%E5%B9%95%E6%88%AA%E5%9B%BE.png",
    "269370": "Cannot agree more with you. My best single model get 0.518 on pv and 0.510 on pb while my best average result get 0.515 on pv but 0.511 on pb. Similar local cv strategy just like other teams but not very helpful. Really a tragedy.\n\n（update: I make a big mistake on my cv set. So I am not sure whether there is a good cv set working for this competition.)",
    "269386": "Can you share how you tried to deal with bias?"
  },
  "source": "meta"
}