{
  "id": 536404,
  "title": "Does it make sense to submit prediction uncertainties below the \"ideal sigma\" of 1e-5?",
  "url": "/competitions/ariel-data-challenge-2024/discussion/536404",
  "author_name": "DennisSakva",
  "post_date": "2024-09-27T09:51:33.166000",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi, everyone!<br>\nI'm trying to understand what's the impact of \"ideal sigma\" on our predictions.<br>\nPurely mechanistically the score continues to improve the closer individual sigmas are to real ones, even if it is below the \"ideal sigma\". </p>",
  "messages": [
    {
      "id": 3000112,
      "postDate": "2024-09-27T10:03:03.290Z",
      "content": "<p>Sure. for some planets, you can make a fairly accurate prediction and a small sigma, say, 5e-6, may well be there. and this increases the total score. but, of course, the average is unlikely to be lower than 1e-5, because everything is drowned in noise.</p>",
      "rawMarkdown": "Sure. for some planets, you can make a fairly accurate prediction and a small sigma, say, 5e-6, may well be there. and this increases the total score. but, of course, the average is unlikely to be lower than 1e-5, because everything is drowned in noise.",
      "votes": 3,
      "replies": [
        {
          "id": 3000295,
          "postDate": "2024-09-27T13:50:48.223Z",
          "content": "<p>Indeed; there is nothing special or 'ideal' about the 1e-5 number. It's just the best the organizers consider reasonably achievable (and I think also what the ARIEL project eventually hopes to achieve).</p>",
          "rawMarkdown": "Indeed; there is nothing special or 'ideal' about the 1e-5 number. It's just the best the organizers consider reasonably achievable (and I think also what the ARIEL project eventually hopes to achieve).",
          "votes": 6
        }
      ]
    },
    {
      "id": 3000110,
      "postDate": "2024-09-27T09:51:33.167Z",
      "content": "<p>Hi, everyone!<br>\nI'm trying to understand what's the impact of \"ideal sigma\" on our predictions.<br>\nPurely mechanistically the score continues to improve the closer individual sigmas are to real ones, even if it is below the \"ideal sigma\". </p>",
      "rawMarkdown": "Hi, everyone!\nI'm trying to understand what's the impact of \"ideal sigma\" on our predictions.\nPurely mechanistically the score continues to improve the closer individual sigmas are to real ones, even if it is below the \"ideal sigma\". ",
      "votes": 4
    },
    {
      "id": 3000818,
      "postDate": "2024-09-28T05:36:19.097Z",
      "content": "<p>I didn't find much difference.</p>",
      "rawMarkdown": "I didn't find much difference.",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 3000112,
      "author_name": "Sergei Fironov",
      "author_url": "",
      "post_date": "2024-09-27T10:03:03.290000",
      "content": "<p>Sure. for some planets, you can make a fairly accurate prediction and a small sigma, say, 5e-6, may well be there. and this increases the total score. but, of course, the average is unlikely to be lower than 1e-5, because everything is drowned in noise.</p>",
      "votes": 3,
      "replies": [
        {
          "id": 3000295,
          "author_name": "Jeroen Cottaar",
          "author_url": "",
          "post_date": "2024-09-27T13:50:48.223000",
          "content": "<p>Indeed; there is nothing special or 'ideal' about the 1e-5 number. It's just the best the organizers consider reasonably achievable (and I think also what the ARIEL project eventually hopes to achieve).</p>",
          "votes": 6,
          "replies": []
        }
      ]
    },
    {
      "id": 3000818,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-09-28T05:36:19.097000",
      "content": "<p>I didn't find much difference.</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3000112": "Sure. for some planets, you can make a fairly accurate prediction and a small sigma, say, 5e-6, may well be there. and this increases the total score. but, of course, the average is unlikely to be lower than 1e-5, because everything is drowned in noise.",
    "3000110": "Hi, everyone!\nI'm trying to understand what's the impact of \"ideal sigma\" on our predictions.\nPurely mechanistically the score continues to improve the closer individual sigmas are to real ones, even if it is below the \"ideal sigma\". ",
    "3000818": "I didn't find much difference."
  }
}