{
  "id": 27535,
  "title": "high scores of new joiners ",
  "url": "/competitions/outbrain-click-prediction/discussion/27535",
  "author_name": "",
  "post_date": "2017-01-10T05:45:14.587Z",
  "votes": 2,
  "comment_count": 6,
  "views": 560,
  "content": "<p>Hey there, </p>\n\n<p>Wondering about your thoughts regarding about new joiners, who register on kaggle like 4-8 days ago, and get places in top 10, and now - the first row. </p>\n\n<p>What could be the nature of such great results? </p>\n\n<p>Thanks,\nAndrii</p>",
  "messages": [
    {
      "id": "155211",
      "postDate": "01/10/2017 05:45:14",
      "content": "<p>Hey there, </p>\n\n<p>Wondering about your thoughts regarding about new joiners, who register on kaggle like 4-8 days ago, and get places in top 10, and now - the first row. </p>\n\n<p>What could be the nature of such great results? </p>\n\n<p>Thanks,\nAndrii</p>",
      "rawMarkdown": "Hey there, \r\n\r\nWondering about your thoughts regarding about new joiners, who register on kaggle like 4-8 days ago, and get places in top 10, and now - the first row. \r\n\r\nWhat could be the nature of such great results? \r\n\r\nThanks,\r\nAndrii",
      "votes": null
    },
    {
      "id": "155212",
      "postDate": "01/10/2017 05:48:34",
      "content": "<p>I wonder exactly the same. :)</p>\n\n<p>We put in a loooooooooooooot of effort to get here and I don't think we can improve significantly with this score but it seems they found something else.</p>",
      "rawMarkdown": "I wonder exactly the same. :)\r\n\r\nWe put in a loooooooooooooot of effort to get here and I don't think we can improve significantly with this score but it seems they found something else.",
      "votes": null
    },
    {
      "id": "155213",
      "postDate": "01/10/2017 05:51:39",
      "content": "<p>that is why:</p>\n\n<p><a href=\"http://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf\">http://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf</a></p>",
      "rawMarkdown": "that is why:\r\n\r\nhttp://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf",
      "votes": null
    },
    {
      "id": "155215",
      "postDate": "01/10/2017 05:52:51",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "155216",
      "postDate": "01/10/2017 05:55:07",
      "content": "<p>I think someone said elsewhere in the forum that if they are from that course, their team name would start with NTU</p>",
      "rawMarkdown": "I think someone said elsewhere in the forum that if they are from that course, their team name would start with NTU",
      "votes": null
    },
    {
      "id": "155220",
      "postDate": "01/10/2017 06:04:20",
      "content": "<p>People from NTU are just that good at click prediction </p>",
      "rawMarkdown": "People from NTU are just that good at click prediction",
      "votes": null
    },
    {
      "id": "155243",
      "postDate": "01/10/2017 08:49:16",
      "content": "<p>It's nice to see many kagglers' discussion. I have done Outbrain Competition for nearly three months, and I keep trying small data offline experiment for CV improve to accelerate the experiment iteration. In the last period my team mate and I put more effort, and also transfer and keep merging our offline technique to lb result using all data.</p>\n\n<p>I think ctr prediction is complex and really coooool, our score still has large room to the perfect ctr algorithm. Hope we could have good luck and we could also discuss about algorithm after the competition finished:-)</p>\n\n<p>Thanks,\nnomo</p>",
      "rawMarkdown": "It's nice to see many kagglers' discussion. I have done Outbrain Competition for nearly three months, and I keep trying small data offline experiment for CV improve to accelerate the experiment iteration. In the last period my team mate and I put more effort, and also transfer and keep merging our offline technique to lb result using all data.\r\n\r\nI think ctr prediction is complex and really coooool, our score still has large room to the perfect ctr algorithm. Hope we could have good luck and we could also discuss about algorithm after the competition finished:-)\r\n\r\nThanks,\r\nnomo",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 155212,
      "author_name": "xiaozhouwang",
      "author_url": "",
      "post_date": "01/10/2017 05:48:34",
      "content": "<p>I wonder exactly the same. :)</p>\n\n<p>We put in a loooooooooooooot of effort to get here and I don't think we can improve significantly with this score but it seems they found something else.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 155213,
      "author_name": "yoshscl",
      "author_url": "",
      "post_date": "01/10/2017 05:51:39",
      "content": "<p>that is why:</p>\n\n<p><a href=\"http://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf\">http://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 155215,
      "author_name": "yoshscl",
      "author_url": "",
      "post_date": "01/10/2017 05:52:51",
      "content": "",
      "votes": null,
      "replies": []
    },
    {
      "id": 155216,
      "author_name": "xiaozhouwang",
      "author_url": "",
      "post_date": "01/10/2017 05:55:07",
      "content": "<p>I think someone said elsewhere in the forum that if they are from that course, their team name would start with NTU</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 155220,
      "author_name": "agrigorev",
      "author_url": "",
      "post_date": "01/10/2017 06:04:20",
      "content": "<p>People from NTU are just that good at click prediction </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 155243,
      "author_name": "ecow1338",
      "author_url": "",
      "post_date": "01/10/2017 08:49:16",
      "content": "<p>It's nice to see many kagglers' discussion. I have done Outbrain Competition for nearly three months, and I keep trying small data offline experiment for CV improve to accelerate the experiment iteration. In the last period my team mate and I put more effort, and also transfer and keep merging our offline technique to lb result using all data.</p>\n\n<p>I think ctr prediction is complex and really coooool, our score still has large room to the perfect ctr algorithm. Hope we could have good luck and we could also discuss about algorithm after the competition finished:-)</p>\n\n<p>Thanks,\nnomo</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "155211": "Hey there, \r\n\r\nWondering about your thoughts regarding about new joiners, who register on kaggle like 4-8 days ago, and get places in top 10, and now - the first row. \r\n\r\nWhat could be the nature of such great results? \r\n\r\nThanks,\r\nAndrii",
    "155212": "I wonder exactly the same. :)\r\n\r\nWe put in a loooooooooooooot of effort to get here and I don't think we can improve significantly with this score but it seems they found something else.",
    "155213": "that is why:\r\n\r\nhttp://speech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/project2.pdf",
    "155215": "",
    "155216": "I think someone said elsewhere in the forum that if they are from that course, their team name would start with NTU",
    "155220": "People from NTU are just that good at click prediction",
    "155243": "It's nice to see many kagglers' discussion. I have done Outbrain Competition for nearly three months, and I keep trying small data offline experiment for CV improve to accelerate the experiment iteration. In the last period my team mate and I put more effort, and also transfer and keep merging our offline technique to lb result using all data.\r\n\r\nI think ctr prediction is complex and really coooool, our score still has large room to the perfect ctr algorithm. Hope we could have good luck and we could also discuss about algorithm after the competition finished:-)\r\n\r\nThanks,\r\nnomo"
  },
  "source": "meta"
}