{
  "id": 16228,
  "title": "A question to the competition host",
  "url": "/competitions/flavours-of-physics/discussion/16228",
  "author_name": "",
  "post_date": "2015-08-31T06:38:57.313Z",
  "votes": null,
  "comment_count": 3,
  "views": 570,
  "content": "<p>I know that the current leader board is calculated on approximately 30% of the test data. So what about the tests? My result has to pass the agreement and correlation test when I submit it to Kaggle every time. Do the two tests base on 30% or 100% of the test data? If only 30% of the data is used,  I am not sure whether the left 70% of the test data can pass the tests. </p>\n\n<p>Thank you for your time.</p>",
  "messages": [
    {
      "id": "90998",
      "postDate": "08/31/2015 06:38:57",
      "content": "<p>I know that the current leader board is calculated on approximately 30% of the test data. So what about the tests? My result has to pass the agreement and correlation test when I submit it to Kaggle every time. Do the two tests base on 30% or 100% of the test data? If only 30% of the data is used,  I am not sure whether the left 70% of the test data can pass the tests. </p>\n\n<p>Thank you for your time.</p>",
      "rawMarkdown": "I know that the current leader board is calculated on approximately 30% of the test data. So what about the tests? My result has to pass the agreement and correlation test when I submit it to Kaggle every time. Do the two tests base on 30% or 100% of the test data? If only 30% of the data is used,  I am not sure whether the left 70% of the test data can pass the tests. \r\n\r\nThank you for your time.",
      "votes": null
    },
    {
      "id": "91049",
      "postDate": "08/31/2015 16:19:34",
      "content": "<p>You can do these tests yourself, see <a href=\"https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb\">https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb</a></p>",
      "rawMarkdown": "You can do these tests yourself, see https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb",
      "votes": null
    },
    {
      "id": "91091",
      "postDate": "08/31/2015 20:32:29",
      "content": "<p>@Liucheng,</p>\n\n<p>The leaderboard is calculated based on 30% of the test data that's &quot;not ignored&quot;, which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. </p>\n\n<p>In other words, if you see that you passed the tests, you passed the tests. </p>\n\n<p>For &quot;doing the tests yourself&quot;, the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. </p>",
      "rawMarkdown": "Liucheng,\r\n\r\nThe leaderboard is calculated based on 30% of the test data that's \"not ignored\", which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. \r\n\r\nIn other words, if you see that you passed the tests, you passed the tests. \r\n\r\nFor \"doing the tests yourself\", the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores.",
      "votes": null
    },
    {
      "id": "91543",
      "postDate": "09/04/2015 04:01:41",
      "content": "<p>Hi Wendy, thanks for your explanation.</p>\n\n<p>[quote=Wendy Kan;91091]</p>\n\n<p>@Liucheng,</p>\n\n<p>The leaderboard is calculated based on 30% of the test data that's &quot;not ignored&quot;, which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. </p>\n\n<p>In other words, if you see that you passed the tests, you passed the tests. </p>\n\n<p>For &quot;doing the tests yourself&quot;, the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. </p>\n\n<p>[/quote]</p>",
      "rawMarkdown": "Hi Wendy, thanks for your explanation.\r\n\r\n[quote=Wendy Kan;91091]\r\n\r\n@Liucheng,\r\n\r\nThe leaderboard is calculated based on 30% of the test data that's \"not ignored\", which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. \r\n\r\nIn other words, if you see that you passed the tests, you passed the tests. \r\n\r\nFor \"doing the tests yourself\", the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. \r\n\r\n[/quote]",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 91049,
      "author_name": "hekkon",
      "author_url": "",
      "post_date": "08/31/2015 16:19:34",
      "content": "<p>You can do these tests yourself, see <a href=\"https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb\">https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 91091,
      "author_name": "wendykan",
      "author_url": "",
      "post_date": "08/31/2015 20:32:29",
      "content": "<p>@Liucheng,</p>\n\n<p>The leaderboard is calculated based on 30% of the test data that's &quot;not ignored&quot;, which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. </p>\n\n<p>In other words, if you see that you passed the tests, you passed the tests. </p>\n\n<p>For &quot;doing the tests yourself&quot;, the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 91543,
      "author_name": "liuchengwang",
      "author_url": "",
      "post_date": "09/04/2015 04:01:41",
      "content": "<p>Hi Wendy, thanks for your explanation.</p>\n\n<p>[quote=Wendy Kan;91091]</p>\n\n<p>@Liucheng,</p>\n\n<p>The leaderboard is calculated based on 30% of the test data that's &quot;not ignored&quot;, which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. </p>\n\n<p>In other words, if you see that you passed the tests, you passed the tests. </p>\n\n<p>For &quot;doing the tests yourself&quot;, the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. </p>\n\n<p>[/quote]</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "90998": "I know that the current leader board is calculated on approximately 30% of the test data. So what about the tests? My result has to pass the agreement and correlation test when I submit it to Kaggle every time. Do the two tests base on 30% or 100% of the test data? If only 30% of the data is used,  I am not sure whether the left 70% of the test data can pass the tests. \r\n\r\nThank you for your time.",
    "91049": "You can do these tests yourself, see https://github.com/yandexdataschool/flavours-of-physics-start/blob/master/flatness_boosting.ipynb",
    "91091": "Liucheng,\r\n\r\nThe leaderboard is calculated based on 30% of the test data that's \"not ignored\", which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. \r\n\r\nIn other words, if you see that you passed the tests, you passed the tests. \r\n\r\nFor \"doing the tests yourself\", the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores.",
    "91543": "Hi Wendy, thanks for your explanation.\r\n\r\n[quote=Wendy Kan;91091]\r\n\r\n@Liucheng,\r\n\r\nThe leaderboard is calculated based on 30% of the test data that's \"not ignored\", which means the agreement test and correlation test are passed, and then these rows are then ignored for scoring. The rest of the test data rows are then used to calculate your score, and you only see 30% of that on LB. \r\n\r\nIn other words, if you see that you passed the tests, you passed the tests. \r\n\r\nFor \"doing the tests yourself\", the data you're given is only a subset of the entire agreement/correlation test data. So you may or may not pass the test if you don't submit it to Kaggle. I will suggest you test locally first, and then submit to Kaggle to make sure you pass the tests and get the scores. \r\n\r\n[/quote]"
  },
  "source": "meta"
}