I saw this sentence in this paper.
\n\n\nThe stocks in a market canshare information so that their stock prices are highly correlated.
\n
We can consider using yesterday's data from other stocks to predict today's stocks.
", "messages": [ { "id": 3037662, "postDate": "2024-11-06T00:23:01.500Z", "content": "I saw this sentence in this paper.
\n\n\nThe stocks in a market canshare information so that their stock prices are highly correlated.
\n
We can consider using yesterday's data from other stocks to predict today's stocks.
", "rawMarkdown": "I saw this sentence in this paper.\n>The stocks in a market canshare information so that their stock prices are highly correlated.\n\nWe can consider using yesterday's data from other stocks to predict today's stocks.", "votes": 9 }, { "id": 3038226, "postDate": "2024-11-06T18:15:56.633Z", "content": "I've checked the correlations between the symbols and it's not that high in general. There are some groups with high-ish correlation (0.60 or so) but most symbols move very uniquely. So I'm not sure we can use them like that.
\nI published it here: https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/544733
This concept is not new, portfolio theory has embedded these concepts since Modern Portfolio Theory (Markowitz).
\nSince we have multiple symbol_ids, we probably have correlated assets. I think (and have started to try) that we can use multiple models for groups of stocks. Sure, we can have one model for each symbol_id, but it will be too complex and/or heavily time-consuming.
", "rawMarkdown": "This concept is not new, portfolio theory has embedded these concepts since Modern Portfolio Theory (Markowitz).\n\nSince we have multiple symbol_ids, we probably have correlated assets. I think (and have started to try) that we can use multiple models for groups of stocks. Sure, we can have one model for each symbol_id, but it will be too complex and/or heavily time-consuming.", "votes": 1, "replies": [ { "id": 3037966, "postDate": "2024-11-06T12:24:56.973Z", "content": "I started this path and have abandoned. There was too much work involved padding out all the input data to handle days with missing symbol_ids and also potential for new symbol_ids. In the end decided to not go with 1 model per symbol accept as potentially a late in the game exploration.
", "rawMarkdown": "I started this path and have abandoned. There was too much work involved padding out all the input data to handle days with missing symbol_ids and also potential for new symbol_ids. In the end decided to not go with 1 model per symbol accept as potentially a late in the game exploration." }, { "id": 3038325, "postDate": "2024-11-06T20:36:09.453Z", "content": "@nandodmelo new symbols can be added to the test dataset anytime and a symbol may be purged off too. In this case, your model may fail
", "rawMarkdown": "@nandodmelo new symbols can be added to the test dataset anytime and a symbol may be purged off too. In this case, your model may fail " } ] }, { "id": 3040441, "postDate": "2024-11-09T06:17:35.387Z", "content": "There could be ETFs or indices that are tracking a specific sector or even a total stock market. That's the reason why they could be correlated.
", "rawMarkdown": "There could be ETFs or indices that are tracking a specific sector or even a total stock market. That's the reason why they could be correlated." }, { "id": 3040396, "postDate": "2024-11-09T04:25:45.420Z", "content": "thanks for sharing scripts and informations its very easy to understand and learn from posts.
", "rawMarkdown": "thanks for sharing scripts and informations its very easy to understand and learn from posts." }, { "id": 3038172, "postDate": "2024-11-06T17:22:08.593Z", "content": "Based on the sentence in this paper.
\n\n\nmodel the relationship between stocks as dynamic company relation graphs
\n
We'd better consider dynamic relations during test time.
", "rawMarkdown": "Based on the sentence in [this paper](https://arxiv.org/pdf/2305.08740).\n>model the relationship between stocks as dynamic company relation graphs\n\nWe'd better consider dynamic relations during test time." }, { "id": 3037918, "postDate": "2024-11-06T11:16:03.593Z", "content": "Thanks for sharing this insightful paper.
\nIn general some stocks will be correlated with each other and their prices will move with a certain lag. This could also mean that its a trending market, where there is a sustained movement in a certain direction.
", "rawMarkdown": "Thanks for sharing this insightful paper. \n\nIn general some stocks will be correlated with each other and their prices will move with a certain lag. This could also mean that its a trending market, where there is a sustained movement in a certain direction. " }, { "id": 3037936, "postDate": "2024-11-06T11:38:19.770Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 3038226, "author_name": "Natan Labarrère", "author_url": "", "post_date": "2024-11-06T18:15:56.633000", "content": "I've checked the correlations between the symbols and it's not that high in general. There are some groups with high-ish correlation (0.60 or so) but most symbols move very uniquely. So I'm not sure we can use them like that.
\nI published it here: https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/544733
This concept is not new, portfolio theory has embedded these concepts since Modern Portfolio Theory (Markowitz).
\nSince we have multiple symbol_ids, we probably have correlated assets. I think (and have started to try) that we can use multiple models for groups of stocks. Sure, we can have one model for each symbol_id, but it will be too complex and/or heavily time-consuming.
", "votes": 1, "replies": [ { "id": 3037966, "author_name": "Michael Timbs", "author_url": "", "post_date": "2024-11-06T12:24:56.973000", "content": "I started this path and have abandoned. There was too much work involved padding out all the input data to handle days with missing symbol_ids and also potential for new symbol_ids. In the end decided to not go with 1 model per symbol accept as potentially a late in the game exploration.
", "votes": 0, "replies": [] }, { "id": 3038325, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2024-11-06T20:36:09.453000", "content": "@nandodmelo new symbols can be added to the test dataset anytime and a symbol may be purged off too. In this case, your model may fail
", "votes": 0, "replies": [] } ] }, { "id": 3040441, "author_name": "Gunes Evitan", "author_url": "", "post_date": "2024-11-09T06:17:35.387000", "content": "There could be ETFs or indices that are tracking a specific sector or even a total stock market. That's the reason why they could be correlated.
", "votes": 0, "replies": [] }, { "id": 3040396, "author_name": "Mihir Sutariya", "author_url": "", "post_date": "2024-11-09T04:25:45.420000", "content": "thanks for sharing scripts and informations its very easy to understand and learn from posts.
", "votes": 0, "replies": [] }, { "id": 3038172, "author_name": "Xiang Sheng", "author_url": "", "post_date": "2024-11-06T17:22:08.593000", "content": "Based on the sentence in this paper.
\n\n\nmodel the relationship between stocks as dynamic company relation graphs
\n
We'd better consider dynamic relations during test time.
", "votes": 0, "replies": [] }, { "id": 3037918, "author_name": "SkyLord", "author_url": "", "post_date": "2024-11-06T11:16:03.593000", "content": "Thanks for sharing this insightful paper.
\nIn general some stocks will be correlated with each other and their prices will move with a certain lag. This could also mean that its a trending market, where there is a sustained movement in a certain direction.
", "votes": 0, "replies": [] }, { "id": 3037936, "author_name": "", "author_url": "", "post_date": "2024-11-06T11:38:19.770000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3037662": "I saw this sentence in this paper.\n>The stocks in a market canshare information so that their stock prices are highly correlated.\n\nWe can consider using yesterday's data from other stocks to predict today's stocks.", "3038226": "I've checked the correlations between the symbols and it's not that high in general. There are some groups with high-ish correlation (0.60 or so) but most symbols move very uniquely. So I'm not sure we can use them like that.\nI published it here: https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/544733", "3037956": "This concept is not new, portfolio theory has embedded these concepts since Modern Portfolio Theory (Markowitz).\n\nSince we have multiple symbol_ids, we probably have correlated assets. I think (and have started to try) that we can use multiple models for groups of stocks. Sure, we can have one model for each symbol_id, but it will be too complex and/or heavily time-consuming.", "3040441": "There could be ETFs or indices that are tracking a specific sector or even a total stock market. That's the reason why they could be correlated.", "3040396": "thanks for sharing scripts and informations its very easy to understand and learn from posts.", "3038172": "Based on the sentence in [this paper](https://arxiv.org/pdf/2305.08740).\n>model the relationship between stocks as dynamic company relation graphs\n\nWe'd better consider dynamic relations during test time.", "3037918": "Thanks for sharing this insightful paper. \n\nIn general some stocks will be correlated with each other and their prices will move with a certain lag. This could also mean that its a trending market, where there is a sustained movement in a certain direction. ", "3037936": "" } }