{
  "id": 134742,
  "title": "Has anyone modeled 2016 and 2017?",
  "url": "/competitions/google-cloud-ncaa-march-madness-2020-division-1-womens-tournament/discussion/134742",
  "author_name": "",
  "post_date": "2020-03-10T04:58:25.675863300Z",
  "votes": null,
  "comment_count": 8,
  "views": 0,
  "content": "<p>2018 was he first year this comp was held. Has anyone looked at how their model performed in the seasons prior? \nI'm specifically curious about 2016 and 2017 because I want to establish as baseline. </p>",
  "messages": [
    {
      "id": "767783",
      "postDate": "03/10/2020 04:58:25",
      "content": "<p>2018 was he first year this comp was held. Has anyone looked at how their model performed in the seasons prior? \nI'm specifically curious about 2016 and 2017 because I want to establish as baseline. </p>",
      "rawMarkdown": "2018 was he first year this comp was held. Has anyone looked at how their model performed in the seasons prior? \nI'm specifically curious about 2016 and 2017 because I want to establish as baseline.",
      "votes": null
    },
    {
      "id": "767819",
      "postDate": "03/10/2020 05:56:27",
      "content": "<p>Hi Ben, </p>\n\n<p>the competition is older than that. Here the 2016 <a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2016\">https://www.kaggle.com/c/march-machine-learning-mania-2016</a>\nand 2017 <a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2017\">https://www.kaggle.com/c/march-machine-learning-mania-2017</a> editions</p>\n\n<p>Cheers</p>",
      "rawMarkdown": "Hi Ben, \n\nthe competition is older than that. Here the 2016 https://www.kaggle.com/c/march-machine-learning-mania-2016\nand 2017 https://www.kaggle.com/c/march-machine-learning-mania-2017 editions\n\nCheers",
      "votes": null
    },
    {
      "id": "768009",
      "postDate": "03/10/2020 11:04:18",
      "content": "<p>Definitely older than that! I participated in the inaugural competition back in... 2014.</p>",
      "rawMarkdown": "Definitely older than that! I participated in the inaugural competition back in... 2014.",
      "votes": null
    },
    {
      "id": "768028",
      "postDate": "03/10/2020 11:34:16",
      "content": "<p>I'm only looking for comparisons for the women's tournament.  </p>",
      "rawMarkdown": "I'm only looking for comparisons for the women's tournament.",
      "votes": null
    },
    {
      "id": "768031",
      "postDate": "03/10/2020 11:35:39",
      "content": "<p>Oh sorry, I misread the competition name :)</p>",
      "rawMarkdown": "Oh sorry, I misread the competition name :)",
      "votes": null
    },
    {
      "id": "768466",
      "postDate": "03/10/2020 20:39:44",
      "content": "<p>I think this discussion is pretty solid evidence that the Women's competition didn't get going until 2018. </p>\n\n<p><a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795\">https://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795</a></p>\n\n<p>Edit: Sorry I misread your question and saw the posts below first. I'd be glad to share my 2016/2017 metrics when I get home today!</p>",
      "rawMarkdown": "I think this discussion is pretty solid evidence that the Women's competition didn't get going until 2018. \n\nhttps://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795\n\nEdit: Sorry I misread your question and saw the posts below first. I'd be glad to share my 2016/2017 metrics when I get home today!",
      "votes": null
    },
    {
      "id": "768515",
      "postDate": "03/10/2020 22:42:00",
      "content": "<p>So I'm still tweaking a bit, but my current model would have me in ~30th place in 2018 and ~90th in 2019 if I could take it back in time.</p>\n\n<p>Edit: <a href=\"/btesch\">@btesch</a> wow, I must have been distracted yesterday. Here are the scores:</p>\n\n<p>2016 - 0.4944\n2017 - 0.3758</p>\n\n<p>No tweaking or clipping on my model. Out of curiosity, how did you find yours performing without any specific modifications after training the model? My model is mainly optimized for the men's tourney, but I've found it does way better in tournaments that don't have many upsets. Perhaps I'm not being aggressive enough?</p>",
      "rawMarkdown": "So I'm still tweaking a bit, but my current model would have me in ~30th place in 2018 and ~90th in 2019 if I could take it back in time.\n\nEdit: @btesch wow, I must have been distracted yesterday. Here are the scores:\n\n2016 - 0.4944\n2017 - 0.3758\n\nNo tweaking or clipping on my model. Out of curiosity, how did you find yours performing without any specific modifications after training the model? My model is mainly optimized for the men's tourney, but I've found it does way better in tournaments that don't have many upsets. Perhaps I'm not being aggressive enough?",
      "votes": null
    },
    {
      "id": "768678",
      "postDate": "03/11/2020 04:46:30",
      "content": "<p>I've modeled 2016 and 2017. \n2016: .3746\n2017: .3305</p>\n\n<p>This uses only pretournament data for each year and assumes I would have split one first round 50/50 game 1-0 and 0-1.</p>\n\n<p>In the men's tournament, you know what the target logloss for each of the last six years was. It helps to establish a baseline for comparison when creating your own model. For this tournament, it doesn't exist which is why I created this post.    </p>\n\n<p>So...anyone else?</p>",
      "rawMarkdown": "I've modeled 2016 and 2017. \n2016: .3746\n2017: .3305\n\nThis uses only pretournament data for each year and assumes I would have split one first round 50/50 game 1-0 and 0-1.\n\nIn the men's tournament, you know what the target logloss for each of the last six years was. It helps to establish a baseline for comparison when creating your own model. For this tournament, it doesn't exist which is why I created this post.    \n\nSo...anyone else?",
      "votes": null
    },
    {
      "id": "769550",
      "postDate": "03/12/2020 02:10:06",
      "content": "<p>My model was not good when using the same approach. \nMy experience is that there are very few similarities between the two. I originally assumed little effort would be required to get the women's results where I wanted them. I was completely wrong. </p>",
      "rawMarkdown": "My model was not good when using the same approach. \nMy experience is that there are very few similarities between the two. I originally assumed little effort would be required to get the women's results where I wanted them. I was completely wrong.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 767819,
      "author_name": "lucabasa",
      "author_url": "",
      "post_date": "03/10/2020 05:56:27",
      "content": "<p>Hi Ben, </p>\n\n<p>the competition is older than that. Here the 2016 <a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2016\">https://www.kaggle.com/c/march-machine-learning-mania-2016</a>\nand 2017 <a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2017\">https://www.kaggle.com/c/march-machine-learning-mania-2017</a> editions</p>\n\n<p>Cheers</p>",
      "votes": null,
      "replies": [
        {
          "id": 768009,
          "author_name": "taffeylewis",
          "author_url": "",
          "post_date": "03/10/2020 11:04:18",
          "content": "<p>Definitely older than that! I participated in the inaugural competition back in... 2014.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 768028,
          "author_name": "btesch",
          "author_url": "",
          "post_date": "03/10/2020 11:34:16",
          "content": "<p>I'm only looking for comparisons for the women's tournament.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 768031,
          "author_name": "lucabasa",
          "author_url": "",
          "post_date": "03/10/2020 11:35:39",
          "content": "<p>Oh sorry, I misread the competition name :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 768466,
      "author_name": "rsa013",
      "author_url": "",
      "post_date": "03/10/2020 20:39:44",
      "content": "<p>I think this discussion is pretty solid evidence that the Women's competition didn't get going until 2018. </p>\n\n<p><a href=\"https://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795\">https://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795</a></p>\n\n<p>Edit: Sorry I misread your question and saw the posts below first. I'd be glad to share my 2016/2017 metrics when I get home today!</p>",
      "votes": null,
      "replies": [
        {
          "id": 768515,
          "author_name": "rsa013",
          "author_url": "",
          "post_date": "03/10/2020 22:42:00",
          "content": "<p>So I'm still tweaking a bit, but my current model would have me in ~30th place in 2018 and ~90th in 2019 if I could take it back in time.</p>\n\n<p>Edit: <a href=\"/btesch\">@btesch</a> wow, I must have been distracted yesterday. Here are the scores:</p>\n\n<p>2016 - 0.4944\n2017 - 0.3758</p>\n\n<p>No tweaking or clipping on my model. Out of curiosity, how did you find yours performing without any specific modifications after training the model? My model is mainly optimized for the men's tourney, but I've found it does way better in tournaments that don't have many upsets. Perhaps I'm not being aggressive enough?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 769550,
          "author_name": "btesch",
          "author_url": "",
          "post_date": "03/12/2020 02:10:06",
          "content": "<p>My model was not good when using the same approach. \nMy experience is that there are very few similarities between the two. I originally assumed little effort would be required to get the women's results where I wanted them. I was completely wrong. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 768678,
      "author_name": "btesch",
      "author_url": "",
      "post_date": "03/11/2020 04:46:30",
      "content": "<p>I've modeled 2016 and 2017. \n2016: .3746\n2017: .3305</p>\n\n<p>This uses only pretournament data for each year and assumes I would have split one first round 50/50 game 1-0 and 0-1.</p>\n\n<p>In the men's tournament, you know what the target logloss for each of the last six years was. It helps to establish a baseline for comparison when creating your own model. For this tournament, it doesn't exist which is why I created this post.    </p>\n\n<p>So...anyone else?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "767783": "2018 was he first year this comp was held. Has anyone looked at how their model performed in the seasons prior? \nI'm specifically curious about 2016 and 2017 because I want to establish as baseline.",
    "767819": "Hi Ben, \n\nthe competition is older than that. Here the 2016 https://www.kaggle.com/c/march-machine-learning-mania-2016\nand 2017 https://www.kaggle.com/c/march-machine-learning-mania-2017 editions\n\nCheers",
    "768009": "Definitely older than that! I participated in the inaugural competition back in... 2014.",
    "768028": "I'm only looking for comparisons for the women's tournament.",
    "768031": "Oh sorry, I misread the competition name :)",
    "768466": "I think this discussion is pretty solid evidence that the Women's competition didn't get going until 2018. \n\nhttps://www.kaggle.com/c/march-machine-learning-mania-2017/discussion/30795\n\nEdit: Sorry I misread your question and saw the posts below first. I'd be glad to share my 2016/2017 metrics when I get home today!",
    "768515": "So I'm still tweaking a bit, but my current model would have me in ~30th place in 2018 and ~90th in 2019 if I could take it back in time.\n\nEdit: @btesch wow, I must have been distracted yesterday. Here are the scores:\n\n2016 - 0.4944\n2017 - 0.3758\n\nNo tweaking or clipping on my model. Out of curiosity, how did you find yours performing without any specific modifications after training the model? My model is mainly optimized for the men's tourney, but I've found it does way better in tournaments that don't have many upsets. Perhaps I'm not being aggressive enough?",
    "768678": "I've modeled 2016 and 2017. \n2016: .3746\n2017: .3305\n\nThis uses only pretournament data for each year and assumes I would have split one first round 50/50 game 1-0 and 0-1.\n\nIn the men's tournament, you know what the target logloss for each of the last six years was. It helps to establish a baseline for comparison when creating your own model. For this tournament, it doesn't exist which is why I created this post.    \n\nSo...anyone else?",
    "769550": "My model was not good when using the same approach. \nMy experience is that there are very few similarities between the two. I originally assumed little effort would be required to get the women's results where I wanted them. I was completely wrong."
  },
  "source": "meta"
}