{
  "id": 130834,
  "title": "How many seasons are you looking back? ",
  "url": "/competitions/google-cloud-ncaa-march-madness-2020-division-1-mens-tournament/discussion/130834",
  "author_name": "",
  "post_date": "2020-02-16T16:54:48.606112300Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Some datasets go back to 1985. Massey Ordinals goes back to 2003 but not all systems go back that far. The play by play data sets goes back five years (although they went back to 2010 if you get the data from last year's comp).</p>\n\n<p>My question is how much training data will everyone be using? \nI'm debating using the play by play data for a metric I really like but I don't like only going back 10 years. </p>\n\n<p>Thoughts?  </p>",
  "messages": [
    {
      "id": "747625",
      "postDate": "02/16/2020 16:54:48",
      "content": "<p>Some datasets go back to 1985. Massey Ordinals goes back to 2003 but not all systems go back that far. The play by play data sets goes back five years (although they went back to 2010 if you get the data from last year's comp).</p>\n\n<p>My question is how much training data will everyone be using? \nI'm debating using the play by play data for a metric I really like but I don't like only going back 10 years. </p>\n\n<p>Thoughts?  </p>",
      "rawMarkdown": "Some datasets go back to 1985. Massey Ordinals goes back to 2003 but not all systems go back that far. The play by play data sets goes back five years (although they went back to 2010 if you get the data from last year's comp).\n\nMy question is how much training data will everyone be using? \nI'm debating using the play by play data for a metric I really like but I don't like only going back 10 years. \n\nThoughts?",
      "votes": null
    },
    {
      "id": "747896",
      "postDate": "02/17/2020 01:30:37",
      "content": "<p>I'm currently training my models with all the data, but year after year, the oldest data will be less relevant as the team get better or drop performance; I will start building some lag features soon to see if there is any positive impact.</p>",
      "rawMarkdown": "I'm currently training my models with all the data, but year after year, the oldest data will be less relevant as the team get better or drop performance; I will start building some lag features soon to see if there is any positive impact.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 747896,
      "author_name": "cv13j0",
      "author_url": "",
      "post_date": "02/17/2020 01:30:37",
      "content": "<p>I'm currently training my models with all the data, but year after year, the oldest data will be less relevant as the team get better or drop performance; I will start building some lag features soon to see if there is any positive impact.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "747625": "Some datasets go back to 1985. Massey Ordinals goes back to 2003 but not all systems go back that far. The play by play data sets goes back five years (although they went back to 2010 if you get the data from last year's comp).\n\nMy question is how much training data will everyone be using? \nI'm debating using the play by play data for a metric I really like but I don't like only going back 10 years. \n\nThoughts?",
    "747896": "I'm currently training my models with all the data, but year after year, the oldest data will be less relevant as the team get better or drop performance; I will start building some lag features soon to see if there is any positive impact."
  },
  "source": "meta"
}