{
  "id": 346578,
  "title": "What is it that the top performers are juicing out of?",
  "url": "/competitions/amex-default-prediction/discussion/346578",
  "author_name": "",
  "post_date": "2022-08-20T10:22:02.775758Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>We all did assume that 0.799 was the highest we were going to reach then 0.800 came and when it too got saturated we reached 0.801.<br>\nWhat exactly do we work on to improve the score. We just see no way further now. Any direction to the solution?</p>",
  "messages": [
    {
      "id": "1906908",
      "postDate": "08/20/2022 10:22:02",
      "content": "<p>We all did assume that 0.799 was the highest we were going to reach then 0.800 came and when it too got saturated we reached 0.801.<br>\nWhat exactly do we work on to improve the score. We just see no way further now. Any direction to the solution?</p>",
      "rawMarkdown": "We all did assume that 0.799 was the highest we were going to reach then 0.800 came and when it too got saturated we reached 0.801.\nWhat exactly do we work on to improve the score. We just see no way further now. Any direction to the solution?",
      "votes": null
    },
    {
      "id": "1906924",
      "postDate": "08/20/2022 10:42:56",
      "content": "<p>We have 0.802 also, just saw the topper's score went up further, even I wonder the same </p>",
      "rawMarkdown": "We have 0.802 also, just saw the topper's score went up further, even I wonder the same",
      "votes": null
    },
    {
      "id": "1906939",
      "postDate": "08/20/2022 11:09:02",
      "content": "<p>There seems to be only one 0.802</p>",
      "rawMarkdown": "There seems to be only one 0.802",
      "votes": null
    },
    {
      "id": "1906964",
      "postDate": "08/20/2022 11:25:20",
      "content": "<p>It's a logarithmic curve, extracting juice after saturation takes lot of experimentation, trial and errors and toying with wild ideas - data augmentation, different models, different set of features, different aggregation, different objectives, different seeds … and the list goes on and on. I believe top teams are also doing no different magic than these.   </p>",
      "rawMarkdown": "It's a logarithmic curve, extracting juice after saturation takes lot of experimentation, trial and errors and toying with wild ideas - data augmentation, different models, different set of features, different aggregation, different objectives, different seeds ... and the list goes on and on. I believe top teams are also doing no different magic than these.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1906924,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "08/20/2022 10:42:56",
      "content": "<p>We have 0.802 also, just saw the topper's score went up further, even I wonder the same </p>",
      "votes": null,
      "replies": [
        {
          "id": 1906939,
          "author_name": "bhavesjain",
          "author_url": "",
          "post_date": "08/20/2022 11:09:02",
          "content": "<p>There seems to be only one 0.802</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1906964,
      "author_name": "tarique7",
      "author_url": "",
      "post_date": "08/20/2022 11:25:20",
      "content": "<p>It's a logarithmic curve, extracting juice after saturation takes lot of experimentation, trial and errors and toying with wild ideas - data augmentation, different models, different set of features, different aggregation, different objectives, different seeds … and the list goes on and on. I believe top teams are also doing no different magic than these.   </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1906908": "We all did assume that 0.799 was the highest we were going to reach then 0.800 came and when it too got saturated we reached 0.801.\nWhat exactly do we work on to improve the score. We just see no way further now. Any direction to the solution?",
    "1906924": "We have 0.802 also, just saw the topper's score went up further, even I wonder the same",
    "1906939": "There seems to be only one 0.802",
    "1906964": "It's a logarithmic curve, extracting juice after saturation takes lot of experimentation, trial and errors and toying with wild ideas - data augmentation, different models, different set of features, different aggregation, different objectives, different seeds ... and the list goes on and on. I believe top teams are also doing no different magic than these."
  },
  "source": "meta"
}