{
  "id": 614317,
  "title": "No overlapping between top ranks",
  "url": "/competitions/MABe-mouse-behavior-detection/discussion/614317",
  "author_name": "",
  "post_date": "2025-11-03T05:49:22.975716500Z",
  "votes": 8,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I did not check the data of this competition yet. But It seems that there is not many overlap between top ranked participants between this competition and NFL competition, although there are many similarities. What could be the reason for this difference?</p>",
  "messages": [
    {
      "id": "3310508",
      "postDate": "11/03/2025 05:49:22",
      "content": "<p>I did not check the data of this competition yet. But It seems that there is not many overlap between top ranked participants between this competition and NFL competition, although there are many similarities. What could be the reason for this difference?</p>",
      "rawMarkdown": "I did not check the data of this competition yet. But It seems that there is not many overlap between top ranked participants between this competition and NFL competition, although there are many similarities. What could be the reason for this difference?",
      "votes": null
    },
    {
      "id": "3310647",
      "postDate": "11/03/2025 10:56:36",
      "content": "<p>I think the reasons might be the following:</p>\n<ol>\n<li>This game has a stable CV , so many top competitors are holding back on submissions to hide their scores.</li>\n<li>Perhaps getting up to speed with a new competition requires more effort?</li>\n</ol>",
      "rawMarkdown": "I think the reasons might be the following:\n1. This game has a stable CV , so many top competitors are holding back on submissions to hide their scores.\n2. Perhaps getting up to speed with a new competition requires more effort?",
      "votes": null
    },
    {
      "id": "3310766",
      "postDate": "11/03/2025 16:04:30",
      "content": "<p>The data quality makes a neural network approach very difficult. If you look at the data, different labs tracks different body parts and use different resolutions, annotations. Incorrect labels from certain labs also makes training anything other than lab-specific models difficult.</p>",
      "rawMarkdown": "The data quality makes a neural network approach very difficult. If you look at the data, different labs tracks different body parts and use different resolutions, annotations. Incorrect labels from certain labs also makes training anything other than lab-specific models difficult.",
      "votes": null
    },
    {
      "id": "3310958",
      "postDate": "11/04/2025 00:56:00",
      "content": "<p>Do you mean CV/LB correlation for this competition is also stable? For NFL I believe there would be a lot of teams trying to behave as if they submit only once, but I do not think it meaningful. Because the better the CV/LB correlation, the less meaningful this \"show\". </p>",
      "rawMarkdown": "Do you mean CV/LB correlation for this competition is also stable? For NFL I believe there would be a lot of teams trying to behave as if they submit only once, but I do not think it meaningful. Because the better the CV/LB correlation, the less meaningful this \"show\".",
      "votes": null
    },
    {
      "id": "3310963",
      "postDate": "11/04/2025 01:02:22",
      "content": "<p>I see, I guess there would be larger shake up here.</p>",
      "rawMarkdown": "I see, I guess there would be larger shake up here.",
      "votes": null
    },
    {
      "id": "3311061",
      "postDate": "11/04/2025 06:23:12",
      "content": "<p>I don't know much about the NFL competition. Is its CV stable as well? If that's the case, I could give it a try.</p>\n<p>And I feel the same way—a low submission count is nothing to show off, unless the hosts decide to factor it into the scoring, lol.</p>",
      "rawMarkdown": "I don't know much about the NFL competition. Is its CV stable as well? If that's the case, I could give it a try.\n\nAnd I feel the same way—a low submission count is nothing to show off, unless the hosts decide to factor it into the scoring, lol.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3310647,
      "author_name": "daylighth",
      "author_url": "",
      "post_date": "11/03/2025 10:56:36",
      "content": "<p>I think the reasons might be the following:</p>\n<ol>\n<li>This game has a stable CV , so many top competitors are holding back on submissions to hide their scores.</li>\n<li>Perhaps getting up to speed with a new competition requires more effort?</li>\n</ol>",
      "votes": null,
      "replies": [
        {
          "id": 3310958,
          "author_name": "yuanzhezhou",
          "author_url": "",
          "post_date": "11/04/2025 00:56:00",
          "content": "<p>Do you mean CV/LB correlation for this competition is also stable? For NFL I believe there would be a lot of teams trying to behave as if they submit only once, but I do not think it meaningful. Because the better the CV/LB correlation, the less meaningful this \"show\". </p>",
          "votes": null,
          "replies": [
            {
              "id": 3311061,
              "author_name": "daylighth",
              "author_url": "",
              "post_date": "11/04/2025 06:23:12",
              "content": "<p>I don't know much about the NFL competition. Is its CV stable as well? If that's the case, I could give it a try.</p>\n<p>And I feel the same way—a low submission count is nothing to show off, unless the hosts decide to factor it into the scoring, lol.</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3310766,
      "author_name": "xiaoz259",
      "author_url": "",
      "post_date": "11/03/2025 16:04:30",
      "content": "<p>The data quality makes a neural network approach very difficult. If you look at the data, different labs tracks different body parts and use different resolutions, annotations. Incorrect labels from certain labs also makes training anything other than lab-specific models difficult.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3310963,
          "author_name": "yuanzhezhou",
          "author_url": "",
          "post_date": "11/04/2025 01:02:22",
          "content": "<p>I see, I guess there would be larger shake up here.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3310508": "I did not check the data of this competition yet. But It seems that there is not many overlap between top ranked participants between this competition and NFL competition, although there are many similarities. What could be the reason for this difference?",
    "3310647": "I think the reasons might be the following:\n1. This game has a stable CV , so many top competitors are holding back on submissions to hide their scores.\n2. Perhaps getting up to speed with a new competition requires more effort?",
    "3310766": "The data quality makes a neural network approach very difficult. If you look at the data, different labs tracks different body parts and use different resolutions, annotations. Incorrect labels from certain labs also makes training anything other than lab-specific models difficult.",
    "3310958": "Do you mean CV/LB correlation for this competition is also stable? For NFL I believe there would be a lot of teams trying to behave as if they submit only once, but I do not think it meaningful. Because the better the CV/LB correlation, the less meaningful this \"show\".",
    "3310963": "I see, I guess there would be larger shake up here.",
    "3311061": "I don't know much about the NFL competition. Is its CV stable as well? If that's the case, I could give it a try.\n\nAnd I feel the same way—a low submission count is nothing to show off, unless the hosts decide to factor it into the scoring, lol."
  },
  "source": "meta"
}