{
  "id": 42577,
  "title": "How does final submission work?",
  "url": "/competitions/favorita-grocery-sales-forecasting/discussion/42577",
  "author_name": "",
  "post_date": "2017-11-01T19:03:19.917714100Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>During the competition, I run my code against the test.csv data, generate a submission file, and submit that file to the web site. It then gets compared to the known (but hidden to me) actual unit_sales data and a score is calculated, and a leaderboard position is consequently established.</p>\n\n<p>What is going to take place during final submission? Are we going to provide our code and it will be run against new private test data to determine final leaderboard standings? Or, Is a new test.csv file going to be provided and then each of us provides 2 submissions of our predicted results? Or what?</p>",
  "messages": [
    {
      "id": "238621",
      "postDate": "11/01/2017 19:03:19",
      "content": "<p>During the competition, I run my code against the test.csv data, generate a submission file, and submit that file to the web site. It then gets compared to the known (but hidden to me) actual unit_sales data and a score is calculated, and a leaderboard position is consequently established.</p>\n\n<p>What is going to take place during final submission? Are we going to provide our code and it will be run against new private test data to determine final leaderboard standings? Or, Is a new test.csv file going to be provided and then each of us provides 2 submissions of our predicted results? Or what?</p>",
      "rawMarkdown": "During the competition, I run my code against the test.csv data, generate a submission file, and submit that file to the web site. It then gets compared to the known (but hidden to me) actual unit_sales data and a score is calculated, and a leaderboard position is consequently established.\n\nWhat is going to take place during final submission? Are we going to provide our code and it will be run against new private test data to determine final leaderboard standings? Or, Is a new test.csv file going to be provided and then each of us provides 2 submissions of our predicted results? Or what?",
      "votes": null
    },
    {
      "id": "238630",
      "postDate": "11/01/2017 19:37:35",
      "content": "<p>I think it's the former one.\nOur kernel code will be used automatically which will be run against the other 50% of the test data\nI don't know about the part where you don't use a kernel</p>",
      "rawMarkdown": "I think it's the former one.\nOur kernel code will be used automatically which will be run against the other 50% of the test data\nI don't know about the part where you don't use a kernel",
      "votes": null
    },
    {
      "id": "238633",
      "postDate": "11/01/2017 19:47:20",
      "content": "<p>Current test.csv already contains both - public and private parts. Visible public LB score is calculated only from public part of test.csv entries. Private LB score is calculated from the other part of test.csv entries, but is left hidden until the very end. So your submissions already contain everything needed for being final submissions, just choose 2 of them (which you believe to be the best) for final scoring before the end.</p>",
      "rawMarkdown": "Current test.csv already contains both - public and private parts. Visible public LB score is calculated only from public part of test.csv entries. Private LB score is calculated from the other part of test.csv entries, but is left hidden until the very end. So your submissions already contain everything needed for being final submissions, just choose 2 of them (which you believe to be the best) for final scoring before the end.",
      "votes": null
    },
    {
      "id": "238634",
      "postDate": "11/01/2017 19:49:31",
      "content": "<p>Okay. Thanks so much for the concise and completely clear answer!</p>",
      "rawMarkdown": "Okay. Thanks so much for the concise and completely clear answer!",
      "votes": null
    },
    {
      "id": "239285",
      "postDate": "11/03/2017 05:38:39",
      "content": "<p>Hi alijs, I think your explanation applies to quite amount of competitions here in Kaggle, about the way public and private LB being handled with submission.  Not sure if there're any other ways, are you aware?</p>",
      "rawMarkdown": "Hi alijs, I think your explanation applies to quite amount of competitions here in Kaggle, about the way public and private LB being handled with submission.  Not sure if there're any other ways, are you aware?",
      "votes": null
    },
    {
      "id": "239582",
      "postDate": "11/03/2017 20:38:25",
      "content": "<p>@DSEverything, Yes, this is how it happens most of the times. Regarding other ways - I've seen also competitions with 2 stages, where final test set is posted only some week before the competition ends.</p>",
      "rawMarkdown": "DSEverything, Yes, this is how it happens most of the times. Regarding other ways - I've seen also competitions with 2 stages, where final test set is posted only some week before the competition ends.",
      "votes": null
    },
    {
      "id": "263564",
      "postDate": "12/30/2017 19:44:31",
      "content": "<p>the test set is split by time, and the public comes first in time  while the private after, correct?</p>",
      "rawMarkdown": "the test set is split by time, and the public comes first in time  while the private after, correct?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 238630,
      "author_name": "uday97",
      "author_url": "",
      "post_date": "11/01/2017 19:37:35",
      "content": "<p>I think it's the former one.\nOur kernel code will be used automatically which will be run against the other 50% of the test data\nI don't know about the part where you don't use a kernel</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 238633,
      "author_name": "alijs1",
      "author_url": "",
      "post_date": "11/01/2017 19:47:20",
      "content": "<p>Current test.csv already contains both - public and private parts. Visible public LB score is calculated only from public part of test.csv entries. Private LB score is calculated from the other part of test.csv entries, but is left hidden until the very end. So your submissions already contain everything needed for being final submissions, just choose 2 of them (which you believe to be the best) for final scoring before the end.</p>",
      "votes": null,
      "replies": [
        {
          "id": 238634,
          "author_name": "gmurrayhays",
          "author_url": "",
          "post_date": "11/01/2017 19:49:31",
          "content": "<p>Okay. Thanks so much for the concise and completely clear answer!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 239285,
          "author_name": "dongxu027",
          "author_url": "",
          "post_date": "11/03/2017 05:38:39",
          "content": "<p>Hi alijs, I think your explanation applies to quite amount of competitions here in Kaggle, about the way public and private LB being handled with submission.  Not sure if there're any other ways, are you aware?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 239582,
          "author_name": "alijs1",
          "author_url": "",
          "post_date": "11/03/2017 20:38:25",
          "content": "<p>@DSEverything, Yes, this is how it happens most of the times. Regarding other ways - I've seen also competitions with 2 stages, where final test set is posted only some week before the competition ends.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 263564,
      "author_name": "pietromarinelli",
      "author_url": "",
      "post_date": "12/30/2017 19:44:31",
      "content": "<p>the test set is split by time, and the public comes first in time  while the private after, correct?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "238621": "During the competition, I run my code against the test.csv data, generate a submission file, and submit that file to the web site. It then gets compared to the known (but hidden to me) actual unit_sales data and a score is calculated, and a leaderboard position is consequently established.\n\nWhat is going to take place during final submission? Are we going to provide our code and it will be run against new private test data to determine final leaderboard standings? Or, Is a new test.csv file going to be provided and then each of us provides 2 submissions of our predicted results? Or what?",
    "238630": "I think it's the former one.\nOur kernel code will be used automatically which will be run against the other 50% of the test data\nI don't know about the part where you don't use a kernel",
    "238633": "Current test.csv already contains both - public and private parts. Visible public LB score is calculated only from public part of test.csv entries. Private LB score is calculated from the other part of test.csv entries, but is left hidden until the very end. So your submissions already contain everything needed for being final submissions, just choose 2 of them (which you believe to be the best) for final scoring before the end.",
    "238634": "Okay. Thanks so much for the concise and completely clear answer!",
    "239285": "Hi alijs, I think your explanation applies to quite amount of competitions here in Kaggle, about the way public and private LB being handled with submission.  Not sure if there're any other ways, are you aware?",
    "239582": "DSEverything, Yes, this is how it happens most of the times. Regarding other ways - I've seen also competitions with 2 stages, where final test set is posted only some week before the competition ends.",
    "263564": "the test set is split by time, and the public comes first in time  while the private after, correct?"
  },
  "source": "meta"
}