{ "id": 322899, "title": "Analysis of the 6th and final update", "url": "/competitions/g-research-crypto-forecasting/discussion/322899", "author_name": "", "post_date": "2022-05-04T08:52:06.052037300Z", "votes": 12, "comment_count": 3, "views": 0, "content": "

Congrats to the winners, and the host for an interesting and great competition.
\nIt was indeed a crypto puzzle finding a good approach ;)

\n

Below we can see the final plots of the LB and model performance.
\nThe top 4 looks robust, will be interesting reading the writeups 👍
\nWe also have some models that jumped +100 and also ~800 places the final update, so it's never over before Kaggle blows the final whistle, that's for sure :)

\n

\"\"
\nTotal LB distribution
\n\"\"
\n\"\"
\n\"\"
\nplus an experimental plot
\n\"\"

", "messages": [ { "id": "1776863", "postDate": "05/04/2022 08:52:06", "content": "

Congrats to the winners, and the host for an interesting and great competition.
\nIt was indeed a crypto puzzle finding a good approach ;)

\n

Below we can see the final plots of the LB and model performance.
\nThe top 4 looks robust, will be interesting reading the writeups 👍
\nWe also have some models that jumped +100 and also ~800 places the final update, so it's never over before Kaggle blows the final whistle, that's for sure :)

\n

\"\"
\nTotal LB distribution
\n\"\"
\n\"\"
\n\"\"
\nplus an experimental plot
\n\"\"

", "rawMarkdown": "Congrats to the winners, and the host for an interesting and great competition.\nIt was indeed a crypto puzzle finding a good approach ;)\n\nBelow we can see the final plots of the LB and model performance.\nThe top 4 looks robust, will be interesting reading the writeups 👍\nWe also have some models that jumped +100 and also ~800 places the final update, so it's never over before Kaggle blows the final whistle, that's for sure :)\n\n![](https://i.postimg.cc/mDdmmS4q/upd101.jpg)\nTotal LB distribution\n![](https://i.postimg.cc/k4MjNp0K/upd102.jpg)\n![](https://i.postimg.cc/mkJ8s2ty/upd103.jpg)\n![](https://i.postimg.cc/KjRJwgvY/upd104.jpg)\nplus an experimental plot \n![](https://i.postimg.cc/0N0WwsB9/upd105.jpg)", "votes": null }, { "id": "1776902", "postDate": "05/04/2022 09:22:35", "content": "

Super impressive by the top 4 teams, Meme Lord Capital in their own league consistently 0.03 <

", "rawMarkdown": "Super impressive by the top 4 teams, Meme Lord Capital in their own league consistently 0.03 <", "votes": null }, { "id": "1777238", "postDate": "05/04/2022 13:47:46", "content": "

It's interesting to see, that with such scarce data of only OHLC there were models who managed to do solid scores that seem non random.
\nI would be curious how these models perform when adding more thorough approaches used in prop firms. Data from the tape, spread between other exchanges, accumulated volume from other exchanges, just to name a few ideas.

", "rawMarkdown": "It's interesting to see, that with such scarce data of only OHLC there were models who managed to do solid scores that seem non random. \nI would be curious how these models perform when adding more thorough approaches used in prop firms. Data from the tape, spread between other exchanges, accumulated volume from other exchanges, just to name a few ideas.", "votes": null }, { "id": "1777717", "postDate": "05/04/2022 18:29:20", "content": "

This comp was an absolute waste of time, and this is why top GMs avoid such competitions.

", "rawMarkdown": "This comp was an absolute waste of time, and this is why top GMs avoid such competitions.", "votes": null } ], "comments": [ { "id": 1776902, "author_name": "julianmukaj", "author_url": "", "post_date": "05/04/2022 09:22:35", "content": "

Super impressive by the top 4 teams, Meme Lord Capital in their own league consistently 0.03 <

", "votes": null, "replies": [] }, { "id": 1777238, "author_name": "rafalradwanski", "author_url": "", "post_date": "05/04/2022 13:47:46", "content": "

It's interesting to see, that with such scarce data of only OHLC there were models who managed to do solid scores that seem non random.
\nI would be curious how these models perform when adding more thorough approaches used in prop firms. Data from the tape, spread between other exchanges, accumulated volume from other exchanges, just to name a few ideas.

", "votes": null, "replies": [] }, { "id": 1777717, "author_name": "azzamradman", "author_url": "", "post_date": "05/04/2022 18:29:20", "content": "

This comp was an absolute waste of time, and this is why top GMs avoid such competitions.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1776863": "Congrats to the winners, and the host for an interesting and great competition.\nIt was indeed a crypto puzzle finding a good approach ;)\n\nBelow we can see the final plots of the LB and model performance.\nThe top 4 looks robust, will be interesting reading the writeups 👍\nWe also have some models that jumped +100 and also ~800 places the final update, so it's never over before Kaggle blows the final whistle, that's for sure :)\n\n![](https://i.postimg.cc/mDdmmS4q/upd101.jpg)\nTotal LB distribution\n![](https://i.postimg.cc/k4MjNp0K/upd102.jpg)\n![](https://i.postimg.cc/mkJ8s2ty/upd103.jpg)\n![](https://i.postimg.cc/KjRJwgvY/upd104.jpg)\nplus an experimental plot \n![](https://i.postimg.cc/0N0WwsB9/upd105.jpg)", "1776902": "Super impressive by the top 4 teams, Meme Lord Capital in their own league consistently 0.03 <", "1777238": "It's interesting to see, that with such scarce data of only OHLC there were models who managed to do solid scores that seem non random. \nI would be curious how these models perform when adding more thorough approaches used in prop firms. Data from the tape, spread between other exchanges, accumulated volume from other exchanges, just to name a few ideas.", "1777717": "This comp was an absolute waste of time, and this is why top GMs avoid such competitions." }, "source": "meta" }