{
  "id": 300170,
  "title": "So... what's \"count\" and it is useful?",
  "url": "/competitions/g-research-crypto-forecasting/discussion/300170",
  "author_name": "",
  "post_date": "2022-01-11T13:32:56.023437300Z",
  "votes": 4,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Dear fellow Kagglers,</p>\n<p>This might be a very trivial question for most people, but I still don't understand the physical meaning of the base feature \"count\". It is officially defined as \"The number of trades that took place this minute\" (<a href=\"https://www.kaggle.com/c/g-research-crypto-forecasting/data\" target=\"_blank\">data page</a>) or \"Total number of trades in the time interval (last minute)\" (<a href=\"https://www.kaggle.com/cstein06/tutorial-to-the-g-research-crypto-competition\" target=\"_blank\">tutorial notebook</a>).</p>\n<p>My questions:</p>\n<ol>\n<li>How does one define a \"trade\" from the definition of \"count\"? How does \"count\" differ from \"volume\"?</li>\n<li>How do you guys use \"count\"? Would you simply feed it into your model or have you found ways to turn it into a component for more complex features?</li>\n</ol>\n<p>Thanks a million! 🙏</p>",
  "messages": [
    {
      "id": "1646105",
      "postDate": "01/11/2022 13:32:56",
      "content": "<p>Dear fellow Kagglers,</p>\n<p>This might be a very trivial question for most people, but I still don't understand the physical meaning of the base feature \"count\". It is officially defined as \"The number of trades that took place this minute\" (<a href=\"https://www.kaggle.com/c/g-research-crypto-forecasting/data\" target=\"_blank\">data page</a>) or \"Total number of trades in the time interval (last minute)\" (<a href=\"https://www.kaggle.com/cstein06/tutorial-to-the-g-research-crypto-competition\" target=\"_blank\">tutorial notebook</a>).</p>\n<p>My questions:</p>\n<ol>\n<li>How does one define a \"trade\" from the definition of \"count\"? How does \"count\" differ from \"volume\"?</li>\n<li>How do you guys use \"count\"? Would you simply feed it into your model or have you found ways to turn it into a component for more complex features?</li>\n</ol>\n<p>Thanks a million! 🙏</p>",
      "rawMarkdown": "Dear fellow Kagglers,\n\nThis might be a very trivial question for most people, but I still don't understand the physical meaning of the base feature \"count\". It is officially defined as \"The number of trades that took place this minute\" ([data page](https://www.kaggle.com/c/g-research-crypto-forecasting/data)) or \"Total number of trades in the time interval (last minute)\" ([tutorial notebook](https://www.kaggle.com/cstein06/tutorial-to-the-g-research-crypto-competition)).\n\nMy questions:\n\n1. How does one define a \"trade\" from the definition of \"count\"? How does \"count\" differ from \"volume\"?\n2. How do you guys use \"count\"? Would you simply feed it into your model or have you found ways to turn it into a component for more complex features?\n\nThanks a million! 🙏",
      "votes": null
    },
    {
      "id": "1647341",
      "postDate": "01/12/2022 14:06:14",
      "content": "<p>Volume is quantity, count is… count. As an example, take AAPL stock. You observe the following trades in a minute: 100 shares at 175.08, 20 shares at 175.10. 100 shares at 175.06. The volume is 220; the count is 3.</p>\n<p>One intuition to help with feature engineering: if you have one sample row with volume of 1000 and count of 1 and one sample row with volume of 1000 and count of 100… what does that mean? In the first sample, you had one large trader; in the second sample you have many small traders. This may be informative about the information content or momentum. As to how to use them… that's the gold in this competition. Given the tight constraints on compute and memory, I believe you can only get away with a small set of features and basic model.</p>",
      "rawMarkdown": "Volume is quantity, count is... count. As an example, take AAPL stock. You observe the following trades in a minute: 100 shares at 175.08, 20 shares at 175.10. 100 shares at 175.06. The volume is 220; the count is 3.\n\nOne intuition to help with feature engineering: if you have one sample row with volume of 1000 and count of 1 and one sample row with volume of 1000 and count of 100... what does that mean? In the first sample, you had one large trader; in the second sample you have many small traders. This may be informative about the information content or momentum. As to how to use them... that's the gold in this competition. Given the tight constraints on compute and memory, I believe you can only get away with a small set of features and basic model.",
      "votes": null
    },
    {
      "id": "1647991",
      "postDate": "01/13/2022 03:35:14",
      "content": "<p>Many thanks for your insight! I'm with you that count measures the number of trades as your example suggests, but it's still not clear to me how a \"trade\" is defined. Consider the following two examples:</p>\n<p>1) Alice places a limit bid order of 100 shares at 175, which is later settled with two counterparties selling 50 shares each. Was there just one trade or two?</p>\n<p>2) Bob places a limit bid order of 100 shares at 175, with 50 shares settled at T=t_1 and the other 50 shares settled at T=t_2. Was there just one trade or two?</p>\n<p>So it just seems to me that a \"trade\" is tricky to define…</p>",
      "rawMarkdown": "Many thanks for your insight! I'm with you that count measures the number of trades as your example suggests, but it's still not clear to me how a \"trade\" is defined. Consider the following two examples:\n\n1) Alice places a limit bid order of 100 shares at 175, which is later settled with two counterparties selling 50 shares each. Was there just one trade or two?\n\n2) Bob places a limit bid order of 100 shares at 175, with 50 shares settled at T=t_1 and the other 50 shares settled at T=t_2. Was there just one trade or two?\n\nSo it just seems to me that a \"trade\" is tricky to define...",
      "votes": null
    },
    {
      "id": "1648719",
      "postDate": "01/13/2022 16:16:30",
      "content": "<p>Crypto markets/exchanges are not regulated I suppose so there may be different reporting conventions. It would be \"standard\" though for: 1) two trades (trades are matched to counterparties, so you have two sellers who need a match); 2) two trades (trades are reported with price, size, and time). </p>",
      "rawMarkdown": "Crypto markets/exchanges are not regulated I suppose so there may be different reporting conventions. It would be \"standard\" though for: 1) two trades (trades are matched to counterparties, so you have two sellers who need a match); 2) two trades (trades are reported with price, size, and time).",
      "votes": null
    },
    {
      "id": "1650241",
      "postDate": "01/14/2022 23:28:30",
      "content": "<p>A trade is a buy or a sell, but i this context, a buy and sell are matched which is a trade (bid/ask matching).</p>\n<p>Hence a count of 3 or 5 means there have been 3 or 5 matches of buy-sell actions (or bid/ask actions). </p>",
      "rawMarkdown": "A trade is a buy or a sell, but i this context, a buy and sell are matched which is a trade (bid/ask matching).\n\nHence a count of 3 or 5 means there have been 3 or 5 matches of buy-sell actions (or bid/ask actions).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1647341,
      "author_name": "marketneutral",
      "author_url": "",
      "post_date": "01/12/2022 14:06:14",
      "content": "<p>Volume is quantity, count is… count. As an example, take AAPL stock. You observe the following trades in a minute: 100 shares at 175.08, 20 shares at 175.10. 100 shares at 175.06. The volume is 220; the count is 3.</p>\n<p>One intuition to help with feature engineering: if you have one sample row with volume of 1000 and count of 1 and one sample row with volume of 1000 and count of 100… what does that mean? In the first sample, you had one large trader; in the second sample you have many small traders. This may be informative about the information content or momentum. As to how to use them… that's the gold in this competition. Given the tight constraints on compute and memory, I believe you can only get away with a small set of features and basic model.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1647991,
          "author_name": "markzhaoty",
          "author_url": "",
          "post_date": "01/13/2022 03:35:14",
          "content": "<p>Many thanks for your insight! I'm with you that count measures the number of trades as your example suggests, but it's still not clear to me how a \"trade\" is defined. Consider the following two examples:</p>\n<p>1) Alice places a limit bid order of 100 shares at 175, which is later settled with two counterparties selling 50 shares each. Was there just one trade or two?</p>\n<p>2) Bob places a limit bid order of 100 shares at 175, with 50 shares settled at T=t_1 and the other 50 shares settled at T=t_2. Was there just one trade or two?</p>\n<p>So it just seems to me that a \"trade\" is tricky to define…</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1648719,
          "author_name": "marketneutral",
          "author_url": "",
          "post_date": "01/13/2022 16:16:30",
          "content": "<p>Crypto markets/exchanges are not regulated I suppose so there may be different reporting conventions. It would be \"standard\" though for: 1) two trades (trades are matched to counterparties, so you have two sellers who need a match); 2) two trades (trades are reported with price, size, and time). </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1650241,
          "author_name": "neomatrix369",
          "author_url": "",
          "post_date": "01/14/2022 23:28:30",
          "content": "<p>A trade is a buy or a sell, but i this context, a buy and sell are matched which is a trade (bid/ask matching).</p>\n<p>Hence a count of 3 or 5 means there have been 3 or 5 matches of buy-sell actions (or bid/ask actions). </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1646105": "Dear fellow Kagglers,\n\nThis might be a very trivial question for most people, but I still don't understand the physical meaning of the base feature \"count\". It is officially defined as \"The number of trades that took place this minute\" ([data page](https://www.kaggle.com/c/g-research-crypto-forecasting/data)) or \"Total number of trades in the time interval (last minute)\" ([tutorial notebook](https://www.kaggle.com/cstein06/tutorial-to-the-g-research-crypto-competition)).\n\nMy questions:\n\n1. How does one define a \"trade\" from the definition of \"count\"? How does \"count\" differ from \"volume\"?\n2. How do you guys use \"count\"? Would you simply feed it into your model or have you found ways to turn it into a component for more complex features?\n\nThanks a million! 🙏",
    "1647341": "Volume is quantity, count is... count. As an example, take AAPL stock. You observe the following trades in a minute: 100 shares at 175.08, 20 shares at 175.10. 100 shares at 175.06. The volume is 220; the count is 3.\n\nOne intuition to help with feature engineering: if you have one sample row with volume of 1000 and count of 1 and one sample row with volume of 1000 and count of 100... what does that mean? In the first sample, you had one large trader; in the second sample you have many small traders. This may be informative about the information content or momentum. As to how to use them... that's the gold in this competition. Given the tight constraints on compute and memory, I believe you can only get away with a small set of features and basic model.",
    "1647991": "Many thanks for your insight! I'm with you that count measures the number of trades as your example suggests, but it's still not clear to me how a \"trade\" is defined. Consider the following two examples:\n\n1) Alice places a limit bid order of 100 shares at 175, which is later settled with two counterparties selling 50 shares each. Was there just one trade or two?\n\n2) Bob places a limit bid order of 100 shares at 175, with 50 shares settled at T=t_1 and the other 50 shares settled at T=t_2. Was there just one trade or two?\n\nSo it just seems to me that a \"trade\" is tricky to define...",
    "1648719": "Crypto markets/exchanges are not regulated I suppose so there may be different reporting conventions. It would be \"standard\" though for: 1) two trades (trades are matched to counterparties, so you have two sellers who need a match); 2) two trades (trades are reported with price, size, and time).",
    "1650241": "A trade is a buy or a sell, but i this context, a buy and sell are matched which is a trade (bid/ask matching).\n\nHence a count of 3 or 5 means there have been 3 or 5 matches of buy-sell actions (or bid/ask actions)."
  },
  "source": "meta"
}