{
  "id": 238924,
  "title": "0.01 discrepancy between LB and notebook score",
  "url": "/competitions/birdclef-2021/discussion/238924",
  "author_name": "",
  "post_date": "2021-05-13T23:56:18.884229900Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>On a couple of notebooks, I noticed that my LB is 0.01 higher than the score displayed in the Notebook Viewer. Does anybody else have the same issue?</p>",
  "messages": [
    {
      "id": "1306622",
      "postDate": "05/13/2021 23:56:18",
      "content": "<p>On a couple of notebooks, I noticed that my LB is 0.01 higher than the score displayed in the Notebook Viewer. Does anybody else have the same issue?</p>",
      "rawMarkdown": "On a couple of notebooks, I noticed that my LB is 0.01 higher than the score displayed in the Notebook Viewer. Does anybody else have the same issue?",
      "votes": null
    },
    {
      "id": "1307932",
      "postDate": "05/14/2021 19:30:48",
      "content": "<p>I would worry if it was the other way around ;)</p>\n<p>More seriously, maybe the kernel version or some dataset version have changed since the public notebook was run.</p>",
      "rawMarkdown": "I would worry if it was the other way around ;)\n\nMore seriously, maybe the kernel version or some dataset version have changed since the public notebook was run.",
      "votes": null
    },
    {
      "id": "1312024",
      "postDate": "05/17/2021 18:59:40",
      "content": "<p>I think it is a rounding issue - I have noticed in the past that on the LB, it rounds 'down', while on the notebook, it rounds 'up'.</p>\n<p>So if your real score is 0.645 you will see 0.65 on LB and and 0.64 on the notebook. Hence you have an additional piece of info that others don't have i.e. 3rd decimal info :)</p>\n<p>Ofcourse, all of this is my hyptothesis and observation.</p>",
      "rawMarkdown": "I think it is a rounding issue - I have noticed in the past that on the LB, it rounds 'down', while on the notebook, it rounds 'up'.\n\nSo if your real score is 0.645 you will see 0.65 on LB and and 0.64 on the notebook. Hence you have an additional piece of info that others don't have i.e. 3rd decimal info :)\n\nOfcourse, all of this is my hyptothesis and observation.",
      "votes": null
    },
    {
      "id": "1312039",
      "postDate": "05/17/2021 19:10:43",
      "content": "<p>LB rounds definitely down</p>",
      "rawMarkdown": "LB rounds definitely down",
      "votes": null
    },
    {
      "id": "1319296",
      "postDate": "05/23/2021 05:18:27",
      "content": "<p>It's the other way around</p>\n<p><img src=\"https://i.imgur.com/kw6M3X4.png\" alt=\"\"></p>\n<p><img src=\"https://i.imgur.com/lb9frXd.png\" alt=\"\"></p>",
      "rawMarkdown": "It's the other way around\n\n![](https://i.imgur.com/kw6M3X4.png)\n\n![](https://i.imgur.com/lb9frXd.png)",
      "votes": null
    },
    {
      "id": "1319389",
      "postDate": "05/23/2021 06:39:15",
      "content": "<p>Yes - looks like a difference in rounding.</p>",
      "rawMarkdown": "Yes - looks like a difference in rounding.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1307932,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "05/14/2021 19:30:48",
      "content": "<p>I would worry if it was the other way around ;)</p>\n<p>More seriously, maybe the kernel version or some dataset version have changed since the public notebook was run.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1312024,
      "author_name": "watzisname",
      "author_url": "",
      "post_date": "05/17/2021 18:59:40",
      "content": "<p>I think it is a rounding issue - I have noticed in the past that on the LB, it rounds 'down', while on the notebook, it rounds 'up'.</p>\n<p>So if your real score is 0.645 you will see 0.65 on LB and and 0.64 on the notebook. Hence you have an additional piece of info that others don't have i.e. 3rd decimal info :)</p>\n<p>Ofcourse, all of this is my hyptothesis and observation.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1312039,
          "author_name": "philippsinger",
          "author_url": "",
          "post_date": "05/17/2021 19:10:43",
          "content": "<p>LB rounds definitely down</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1319296,
          "author_name": "philipkd",
          "author_url": "",
          "post_date": "05/23/2021 05:18:27",
          "content": "<p>It's the other way around</p>\n<p><img src=\"https://i.imgur.com/kw6M3X4.png\" alt=\"\"></p>\n<p><img src=\"https://i.imgur.com/lb9frXd.png\" alt=\"\"></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1319389,
          "author_name": "watzisname",
          "author_url": "",
          "post_date": "05/23/2021 06:39:15",
          "content": "<p>Yes - looks like a difference in rounding.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1306622": "On a couple of notebooks, I noticed that my LB is 0.01 higher than the score displayed in the Notebook Viewer. Does anybody else have the same issue?",
    "1307932": "I would worry if it was the other way around ;)\n\nMore seriously, maybe the kernel version or some dataset version have changed since the public notebook was run.",
    "1312024": "I think it is a rounding issue - I have noticed in the past that on the LB, it rounds 'down', while on the notebook, it rounds 'up'.\n\nSo if your real score is 0.645 you will see 0.65 on LB and and 0.64 on the notebook. Hence you have an additional piece of info that others don't have i.e. 3rd decimal info :)\n\nOfcourse, all of this is my hyptothesis and observation.",
    "1312039": "LB rounds definitely down",
    "1319296": "It's the other way around\n\n![](https://i.imgur.com/kw6M3X4.png)\n\n![](https://i.imgur.com/lb9frXd.png)",
    "1319389": "Yes - looks like a difference in rounding."
  },
  "source": "meta"
}