{
  "id": 545846,
  "title": "How is -7.8 score be possible 😆",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/545846",
  "author_name": "",
  "post_date": "2024-11-12T12:08:06.988778800Z",
  "votes": 5,
  "comment_count": 12,
  "views": 0,
  "content": "<p>I used PCA and Scaler, but the result was mind blowing. -7.8 ? what….<br>\nDo people even do feature enginnering at all? this score is absurd.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4858569%2F5b214714c20bfc40b5c21c61fcfae1a0%2Fwechat_2024-11-12_200513_068.png?generation=1731413204334622&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "3043413",
      "postDate": "11/12/2024 12:08:06",
      "content": "<p>I used PCA and Scaler, but the result was mind blowing. -7.8 ? what….<br>\nDo people even do feature enginnering at all? this score is absurd.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4858569%2F5b214714c20bfc40b5c21c61fcfae1a0%2Fwechat_2024-11-12_200513_068.png?generation=1731413204334622&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "I used PCA and Scaler, but the result was mind blowing. -7.8 ? what....\nDo people even do feature enginnering at all? this score is absurd.\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4858569%2F5b214714c20bfc40b5c21c61fcfae1a0%2Fwechat_2024-11-12_200513_068.png?generation=1731413204334622&alt=media)",
      "votes": null
    },
    {
      "id": "3043428",
      "postDate": "11/12/2024 12:17:29",
      "content": "<p>🤣 what model is it?</p>",
      "rawMarkdown": "🤣 what model is it?",
      "votes": null
    },
    {
      "id": "3043433",
      "postDate": "11/12/2024 12:19:14",
      "content": "<p>You've nailed it! inverse your predictions and the gold is yours 😅</p>",
      "rawMarkdown": "You've nailed it! inverse your predictions and the gold is yours 😅",
      "votes": null
    },
    {
      "id": "3043441",
      "postDate": "11/12/2024 12:24:37",
      "content": "<p>Its a LGB model…  the test set  during my training was ok! it was 0.005 somthing..</p>",
      "rawMarkdown": "Its a LGB model...  the test set  during my training was ok! it was 0.005 somthing..",
      "votes": null
    },
    {
      "id": "3043442",
      "postDate": "11/12/2024 12:24:59",
      "content": "<p>lol, wish its that simple, …but Ill give it a try</p>",
      "rawMarkdown": "lol, wish its that simple, ...but Ill give it a try",
      "votes": null
    },
    {
      "id": "3043470",
      "postDate": "11/12/2024 12:45:02",
      "content": "<p>He can stop working for the rest of the competition and 1st place will still be his 😅</p>",
      "rawMarkdown": "He can stop working for the rest of the competition and 1st place will still be his 😅",
      "votes": null
    },
    {
      "id": "3043482",
      "postDate": "11/12/2024 12:49:45",
      "content": "<p>If we're competing for the lowest score, take a look at my -24 score… 😂</p>",
      "rawMarkdown": "If we're competing for the lowest score, take a look at my -24 score... 😂",
      "votes": null
    },
    {
      "id": "3043490",
      "postDate": "11/12/2024 12:51:52",
      "content": "<p>At least the submission passed without an error</p>",
      "rawMarkdown": "At least the submission passed without an error",
      "votes": null
    },
    {
      "id": "3043504",
      "postDate": "11/12/2024 12:58:36",
      "content": "<p><a href=\"https://www.kaggle.com/yeraypabongonzalez\" target=\"_blank\">@yeraypabongonzalez</a> how does one compete for the <strong>lowest score</strong>?😀</p>",
      "rawMarkdown": "yeraypabongonzalez how does one compete for the **lowest score**?😀",
      "votes": null
    },
    {
      "id": "3043522",
      "postDate": "11/12/2024 13:15:45",
      "content": "<p>what did you do…</p>",
      "rawMarkdown": "what did you do...",
      "votes": null
    },
    {
      "id": "3044395",
      "postDate": "11/13/2024 11:02:36",
      "content": "<p>That may means your the error of your prediction is too large.Because generally speaking, the R-squared score is calculated by 1 - SSE/SST, and the SSE is the residual error between prediction and ground truth.</p>",
      "rawMarkdown": "That may means your the error of your prediction is too large.Because generally speaking, the R-squared score is calculated by 1 - SSE/SST, and the SSE is the residual error between prediction and ground truth.",
      "votes": null
    },
    {
      "id": "3045125",
      "postDate": "11/14/2024 07:32:35",
      "content": "<p>great explaination with formula.😄</p>",
      "rawMarkdown": "great explaination with formula.😄",
      "votes": null
    },
    {
      "id": "3056117",
      "postDate": "11/26/2024 15:17:38",
      "content": "<p>I came across similar bugs early in training. Two key causes in order where</p>\n<ol>\n<li>Bugs leading to NaN predictions which got clipped to -5</li>\n<li>Forgetting to clip at all</li>\n</ol>",
      "rawMarkdown": "I came across similar bugs early in training. Two key causes in order where\n1. Bugs leading to NaN predictions which got clipped to -5\n2. Forgetting to clip at all",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3043428,
      "author_name": "shiyili",
      "author_url": "",
      "post_date": "11/12/2024 12:17:29",
      "content": "<p>🤣 what model is it?</p>",
      "votes": null,
      "replies": [
        {
          "id": 3043441,
          "author_name": "zoutain",
          "author_url": "",
          "post_date": "11/12/2024 12:24:37",
          "content": "<p>Its a LGB model…  the test set  during my training was ok! it was 0.005 somthing..</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3043433,
      "author_name": "eu1234",
      "author_url": "",
      "post_date": "11/12/2024 12:19:14",
      "content": "<p>You've nailed it! inverse your predictions and the gold is yours 😅</p>",
      "votes": null,
      "replies": [
        {
          "id": 3043442,
          "author_name": "zoutain",
          "author_url": "",
          "post_date": "11/12/2024 12:24:59",
          "content": "<p>lol, wish its that simple, …but Ill give it a try</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 3043470,
          "author_name": "aymanallawi",
          "author_url": "",
          "post_date": "11/12/2024 12:45:02",
          "content": "<p>He can stop working for the rest of the competition and 1st place will still be his 😅</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3043482,
      "author_name": "yeraypabongonzalez",
      "author_url": "",
      "post_date": "11/12/2024 12:49:45",
      "content": "<p>If we're competing for the lowest score, take a look at my -24 score… 😂</p>",
      "votes": null,
      "replies": [
        {
          "id": 3043490,
          "author_name": "eu1234",
          "author_url": "",
          "post_date": "11/12/2024 12:51:52",
          "content": "<p>At least the submission passed without an error</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 3043504,
          "author_name": "ravi20076",
          "author_url": "",
          "post_date": "11/12/2024 12:58:36",
          "content": "<p><a href=\"https://www.kaggle.com/yeraypabongonzalez\" target=\"_blank\">@yeraypabongonzalez</a> how does one compete for the <strong>lowest score</strong>?😀</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 3043522,
          "author_name": "zoutain",
          "author_url": "",
          "post_date": "11/12/2024 13:15:45",
          "content": "<p>what did you do…</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3044395,
      "author_name": "i2nfinit3y",
      "author_url": "",
      "post_date": "11/13/2024 11:02:36",
      "content": "<p>That may means your the error of your prediction is too large.Because generally speaking, the R-squared score is calculated by 1 - SSE/SST, and the SSE is the residual error between prediction and ground truth.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3045125,
          "author_name": "zoutain",
          "author_url": "",
          "post_date": "11/14/2024 07:32:35",
          "content": "<p>great explaination with formula.😄</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 3056117,
      "author_name": "michaeltimbs",
      "author_url": "",
      "post_date": "11/26/2024 15:17:38",
      "content": "<p>I came across similar bugs early in training. Two key causes in order where</p>\n<ol>\n<li>Bugs leading to NaN predictions which got clipped to -5</li>\n<li>Forgetting to clip at all</li>\n</ol>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3043413": "I used PCA and Scaler, but the result was mind blowing. -7.8 ? what....\nDo people even do feature enginnering at all? this score is absurd.\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4858569%2F5b214714c20bfc40b5c21c61fcfae1a0%2Fwechat_2024-11-12_200513_068.png?generation=1731413204334622&alt=media)",
    "3043428": "🤣 what model is it?",
    "3043433": "You've nailed it! inverse your predictions and the gold is yours 😅",
    "3043441": "Its a LGB model...  the test set  during my training was ok! it was 0.005 somthing..",
    "3043442": "lol, wish its that simple, ...but Ill give it a try",
    "3043470": "He can stop working for the rest of the competition and 1st place will still be his 😅",
    "3043482": "If we're competing for the lowest score, take a look at my -24 score... 😂",
    "3043490": "At least the submission passed without an error",
    "3043504": "yeraypabongonzalez how does one compete for the **lowest score**?😀",
    "3043522": "what did you do...",
    "3044395": "That may means your the error of your prediction is too large.Because generally speaking, the R-squared score is calculated by 1 - SSE/SST, and the SSE is the residual error between prediction and ground truth.",
    "3045125": "great explaination with formula.😄",
    "3056117": "I came across similar bugs early in training. Two key causes in order where\n1. Bugs leading to NaN predictions which got clipped to -5\n2. Forgetting to clip at all"
  },
  "source": "meta"
}