{
  "id": 207705,
  "title": "Ways to address that a random guess is of 25% chance being the correct answer",
  "url": "/competitions/riiid-test-answer-prediction/discussion/207705",
  "author_name": "",
  "post_date": "2020-12-30T22:53:20.553189500Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I started this competition kinda late and curious about one thing (sorry if someone already answered/asked about this): </p>\n<p>For every question, is the training data prepared in the following way?</p>\n<ul>\n<li>The choice is randomized when a user see them, yet in the training data, it is re-aligned to be uniform. </li>\n</ul>\n<p>For example, question 0 has 4 choices being (0) Apple, (1) Bravo, (2) Charlie, (3) Delta; user 123 sees (0) Apple, (1) Bravo, (2) Charlie, (3) Delta, and chose Charlie. While user 321 sees (0) Bravo, (1) Apple, (2) Delta, (3) Charlie, and chose Charlie. In the training data, the choices for these two users are re-aligned to be (2). </p>\n<ul>\n<li>Fixed order. All users see the same order of choices.</li>\n</ul>",
  "messages": [
    {
      "id": "1133074",
      "postDate": "12/30/2020 22:53:20",
      "content": "<p>I started this competition kinda late and curious about one thing (sorry if someone already answered/asked about this): </p>\n<p>For every question, is the training data prepared in the following way?</p>\n<ul>\n<li>The choice is randomized when a user see them, yet in the training data, it is re-aligned to be uniform. </li>\n</ul>\n<p>For example, question 0 has 4 choices being (0) Apple, (1) Bravo, (2) Charlie, (3) Delta; user 123 sees (0) Apple, (1) Bravo, (2) Charlie, (3) Delta, and chose Charlie. While user 321 sees (0) Bravo, (1) Apple, (2) Delta, (3) Charlie, and chose Charlie. In the training data, the choices for these two users are re-aligned to be (2). </p>\n<ul>\n<li>Fixed order. All users see the same order of choices.</li>\n</ul>",
      "rawMarkdown": "I started this competition kinda late and curious about one thing (sorry if someone already answered/asked about this): \n\nFor every question, is the training data prepared in the following way?\n\n- The choice is randomized when a user see them, yet in the training data, it is re-aligned to be uniform. \n\nFor example, question 0 has 4 choices being (0) Apple, (1) Bravo, (2) Charlie, (3) Delta; user 123 sees (0) Apple, (1) Bravo, (2) Charlie, (3) Delta, and chose Charlie. While user 321 sees (0) Bravo, (1) Apple, (2) Delta, (3) Charlie, and chose Charlie. In the training data, the choices for these two users are re-aligned to be (2). \n\n- Fixed order. All users see the same order of choices.",
      "votes": null
    },
    {
      "id": "1133076",
      "postDate": "12/30/2020 22:56:07",
      "content": "<p>The order they see it in is irrelevant because it's been fixed for us s.t. the choice and the right answer per content_id is constant across the dataset. Also, beware—not all questions have 4 possible responses.</p>",
      "rawMarkdown": "The order they see it in is irrelevant because it's been fixed for us s.t. the choice and the right answer per content_id is constant across the dataset. Also, beware—not all questions have 4 possible responses.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1133076,
      "author_name": "authman",
      "author_url": "",
      "post_date": "12/30/2020 22:56:07",
      "content": "<p>The order they see it in is irrelevant because it's been fixed for us s.t. the choice and the right answer per content_id is constant across the dataset. Also, beware—not all questions have 4 possible responses.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1133074": "I started this competition kinda late and curious about one thing (sorry if someone already answered/asked about this): \n\nFor every question, is the training data prepared in the following way?\n\n- The choice is randomized when a user see them, yet in the training data, it is re-aligned to be uniform. \n\nFor example, question 0 has 4 choices being (0) Apple, (1) Bravo, (2) Charlie, (3) Delta; user 123 sees (0) Apple, (1) Bravo, (2) Charlie, (3) Delta, and chose Charlie. While user 321 sees (0) Bravo, (1) Apple, (2) Delta, (3) Charlie, and chose Charlie. In the training data, the choices for these two users are re-aligned to be (2). \n\n- Fixed order. All users see the same order of choices.",
    "1133076": "The order they see it in is irrelevant because it's been fixed for us s.t. the choice and the right answer per content_id is constant across the dataset. Also, beware—not all questions have 4 possible responses."
  },
  "source": "meta"
}