Dear @drwtrading team, I have some questions about the data. How should I contact you for asking these questions?
", "messages": [ { "id": "3216664", "postDate": "06/03/2025 23:03:24", "content": "Dear @drwtrading team, I have some questions about the data. How should I contact you for asking these questions?
", "rawMarkdown": "Dear @drwtrading team, I have some questions about the data. How should I contact you for asking these questions?", "votes": null }, { "id": "3216694", "postDate": "06/04/2025 02:03:50", "content": "Hi Ding, what makes your score a boost in the LB?
", "rawMarkdown": "Hi Ding, what makes your score a boost in the LB?", "votes": null }, { "id": "3216770", "postDate": "06/04/2025 04:55:21", "content": "You can click on their homepage and use the “Contact” link in the top right to email them. That might be better if you’re unsure about raising the data issue in public.
", "rawMarkdown": "You can click on their homepage and use the “Contact” link in the top right to email them. That might be better if you’re unsure about raising the data issue in public.", "votes": null }, { "id": "3216836", "postDate": "06/04/2025 06:37:19", "content": "getting 2nd place, and kick eveyone else's ass by asking naive questions
", "rawMarkdown": "getting 2nd place, and kick eveyone else's ass by asking naive questions", "votes": null }, { "id": "3216839", "postDate": "06/04/2025 06:43:06", "content": "I don't know the question, but let me answer anyway. :D.
\nThis approach is incorrect. You cannot predict timestamp values in the test set and then derive time features from these predictions to develop a model and improve its score. The fundamental objective here is to develop a model based solely on the training data, without using any temporal information from the test set. Even though predictions are used, this approach does not produce a valid solution for real-world scenarios where we encounter situations with no temporal information available.
", "rawMarkdown": "I don't know the question, but let me answer anyway. :D.\n\nThis approach is incorrect. You cannot predict timestamp values in the test set and then derive time features from these predictions to develop a model and improve its score. The fundamental objective here is to develop a model based solely on the training data, without using any temporal information from the test set. Even though predictions are used, this approach does not produce a valid solution for real-world scenarios where we encounter situations with no temporal information available.", "votes": null }, { "id": "3216953", "postDate": "06/04/2025 10:29:26", "content": "Thanks for your guidance! I'm a new one on kaggle so I don't know lots of stuff😅
", "rawMarkdown": "Thanks for your guidance! I'm a new one on kaggle so I don't know lots of stuff😅", "votes": null }, { "id": "3216954", "postDate": "06/04/2025 10:32:21", "content": "I'm also one of the competitors and I don't want to unintentionally break the rule. 😅
", "rawMarkdown": "I'm also one of the competitors and I don't want to unintentionally break the rule. 😅", "votes": null }, { "id": "3216955", "postDate": "06/04/2025 10:34:56", "content": "Thanks for your answer! I've asked the question to the sponor and hope they can reply me as soon as possible
", "rawMarkdown": "Thanks for your answer! I've asked the question to the sponor and hope they can reply me as soon as possible", "votes": null }, { "id": "3216978", "postDate": "06/04/2025 11:20:57", "content": "Why are you not asking it publicly so everyone can see the response?
", "rawMarkdown": "Why are you not asking it publicly so everyone can see the response?", "votes": null }, { "id": "3217021", "postDate": "06/04/2025 12:27:22", "content": "Just asking a rule about their mention:
\n\"You’re welcome to use all the data we provide and any freely and publicly available external datasets, including pre-trained models, to build your solution. Just keep in mind that, similar to the note about using future data, any external dataset containing future information will be considered a rule violation. So make sure everything you use would have been available at the time of prediction!\"
\nIf I get feedback, I will let you know.
", "rawMarkdown": "Just asking a rule about their mention: \n\n\"You’re welcome to use all the data we provide and any freely and publicly available external datasets, including pre-trained models, to build your solution. Just keep in mind that, similar to the note about using future data, any external dataset containing future information will be considered a rule violation. So make sure everything you use would have been available at the time of prediction!\" \n\nIf I get feedback, I will let you know.", "votes": null }, { "id": "3217158", "postDate": "06/04/2025 15:33:48", "content": "I personally think that the usage about external data produces a lot of troubles.
\nThe timestamp in test data is masked so we cannot apply external data on test data. Even there is someway to figure the timestamp out in test data, it should be against the DRW’s will.
The time zone of the data is unclear so if you use external data wrongly in some way, you are actually using future data.
The time cost and delay to get external data is not considered. If one tried to use like economic data, there will be huge time delay before you can really use the data.
Thanks for your answer, that's helpful! If I receive a reply stating that I violated the rule, I will choose the other submission for my final evaluation.
", "rawMarkdown": "Thanks for your answer, that's helpful! If I receive a reply stating that I violated the rule, I will choose the other submission for my final evaluation.", "votes": null }, { "id": "3217861", "postDate": "06/05/2025 13:13:59", "content": "Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024.
\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.
", "rawMarkdown": "Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024. \n\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.", "votes": null }, { "id": "3217863", "postDate": "06/05/2025 13:14:50", "content": "\n\nJust get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024.
\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.
\n
Please correct me if I'm wrong @drwtrading
", "rawMarkdown": "> Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024. \n> \n> You should pretend that we are now on February 29, 2024 and want to build the model that could predict future.\n\nPlease correct me if I'm wrong @drwtrading", "votes": null }, { "id": "3217867", "postDate": "06/05/2025 13:17:18", "content": "Thanks! That was my understanding as we are trying to predict future. Were you using future dataset?
", "rawMarkdown": "Thanks! That was my understanding as we are trying to predict future. Were you using future dataset?", "votes": null }, { "id": "3224640", "postDate": "06/15/2025 08:28:06", "content": "Any secrets you’re holding back? Or are you just waiting for someone else to overtake your second place while you stay stuck in the same spot?
", "rawMarkdown": "Any secrets you’re holding back? Or are you just waiting for someone else to overtake your second place while you stay stuck in the same spot?", "votes": null } ], "comments": [ { "id": 3216694, "author_name": "abdonson", "author_url": "", "post_date": "06/04/2025 02:03:50", "content": "Hi Ding, what makes your score a boost in the LB?
", "votes": null, "replies": [] }, { "id": 3216770, "author_name": "maaax", "author_url": "", "post_date": "06/04/2025 04:55:21", "content": "You can click on their homepage and use the “Contact” link in the top right to email them. That might be better if you’re unsure about raising the data issue in public.
", "votes": null, "replies": [ { "id": 3216836, "author_name": "abdonson", "author_url": "", "post_date": "06/04/2025 06:37:19", "content": "getting 2nd place, and kick eveyone else's ass by asking naive questions
", "votes": null, "replies": [ { "id": 3216954, "author_name": "dingyangwang", "author_url": "", "post_date": "06/04/2025 10:32:21", "content": "I'm also one of the competitors and I don't want to unintentionally break the rule. 😅
", "votes": null, "replies": [] } ] }, { "id": 3216953, "author_name": "dingyangwang", "author_url": "", "post_date": "06/04/2025 10:29:26", "content": "Thanks for your guidance! I'm a new one on kaggle so I don't know lots of stuff😅
", "votes": null, "replies": [] } ] }, { "id": 3216839, "author_name": "onurkoc83", "author_url": "", "post_date": "06/04/2025 06:43:06", "content": "I don't know the question, but let me answer anyway. :D.
\nThis approach is incorrect. You cannot predict timestamp values in the test set and then derive time features from these predictions to develop a model and improve its score. The fundamental objective here is to develop a model based solely on the training data, without using any temporal information from the test set. Even though predictions are used, this approach does not produce a valid solution for real-world scenarios where we encounter situations with no temporal information available.
", "votes": null, "replies": [ { "id": 3216955, "author_name": "dingyangwang", "author_url": "", "post_date": "06/04/2025 10:34:56", "content": "Thanks for your answer! I've asked the question to the sponor and hope they can reply me as soon as possible
", "votes": null, "replies": [] } ] }, { "id": 3216978, "author_name": "julianmukaj", "author_url": "", "post_date": "06/04/2025 11:20:57", "content": "Why are you not asking it publicly so everyone can see the response?
", "votes": null, "replies": [ { "id": 3217021, "author_name": "dingyangwang", "author_url": "", "post_date": "06/04/2025 12:27:22", "content": "Just asking a rule about their mention:
\n\"You’re welcome to use all the data we provide and any freely and publicly available external datasets, including pre-trained models, to build your solution. Just keep in mind that, similar to the note about using future data, any external dataset containing future information will be considered a rule violation. So make sure everything you use would have been available at the time of prediction!\"
\nIf I get feedback, I will let you know.
", "votes": null, "replies": [ { "id": 3217158, "author_name": "yw2735", "author_url": "", "post_date": "06/04/2025 15:33:48", "content": "I personally think that the usage about external data produces a lot of troubles.
\nThe timestamp in test data is masked so we cannot apply external data on test data. Even there is someway to figure the timestamp out in test data, it should be against the DRW’s will.
The time zone of the data is unclear so if you use external data wrongly in some way, you are actually using future data.
The time cost and delay to get external data is not considered. If one tried to use like economic data, there will be huge time delay before you can really use the data.
Thanks for your answer, that's helpful! If I receive a reply stating that I violated the rule, I will choose the other submission for my final evaluation.
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 3217861, "author_name": "dingyangwang", "author_url": "", "post_date": "06/05/2025 13:13:59", "content": "Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024.
\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.
", "votes": null, "replies": [ { "id": 3217863, "author_name": "dingyangwang", "author_url": "", "post_date": "06/05/2025 13:14:50", "content": "\n\nJust get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024.
\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.
\n
Please correct me if I'm wrong @drwtrading
", "votes": null, "replies": [] }, { "id": 3217867, "author_name": "byunjins", "author_url": "", "post_date": "06/05/2025 13:17:18", "content": "Thanks! That was my understanding as we are trying to predict future. Were you using future dataset?
", "votes": null, "replies": [] } ] }, { "id": 3224640, "author_name": "abdonson", "author_url": "", "post_date": "06/15/2025 08:28:06", "content": "Any secrets you’re holding back? Or are you just waiting for someone else to overtake your second place while you stay stuck in the same spot?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3216664": "Dear @drwtrading team, I have some questions about the data. How should I contact you for asking these questions?", "3216694": "Hi Ding, what makes your score a boost in the LB?", "3216770": "You can click on their homepage and use the “Contact” link in the top right to email them. That might be better if you’re unsure about raising the data issue in public.", "3216836": "getting 2nd place, and kick eveyone else's ass by asking naive questions", "3216839": "I don't know the question, but let me answer anyway. :D.\n\nThis approach is incorrect. You cannot predict timestamp values in the test set and then derive time features from these predictions to develop a model and improve its score. The fundamental objective here is to develop a model based solely on the training data, without using any temporal information from the test set. Even though predictions are used, this approach does not produce a valid solution for real-world scenarios where we encounter situations with no temporal information available.", "3216953": "Thanks for your guidance! I'm a new one on kaggle so I don't know lots of stuff😅", "3216954": "I'm also one of the competitors and I don't want to unintentionally break the rule. 😅", "3216955": "Thanks for your answer! I've asked the question to the sponor and hope they can reply me as soon as possible", "3216978": "Why are you not asking it publicly so everyone can see the response?", "3217021": "Just asking a rule about their mention: \n\n\"You’re welcome to use all the data we provide and any freely and publicly available external datasets, including pre-trained models, to build your solution. Just keep in mind that, similar to the note about using future data, any external dataset containing future information will be considered a rule violation. So make sure everything you use would have been available at the time of prediction!\" \n\nIf I get feedback, I will let you know.", "3217158": "I personally think that the usage about external data produces a lot of troubles. \n\n1. The timestamp in test data is masked so we cannot apply external data on test data. Even there is someway to figure the timestamp out in test data, it should be against the DRW’s will.\n\n2. The time zone of the data is unclear so if you use external data wrongly in some way, you are actually using future data.\n\n3. The time cost and delay to get external data is not considered. If one tried to use like economic data, there will be huge time delay before you can really use the data.", "3217200": "Thanks for your answer, that's helpful! If I receive a reply stating that I violated the rule, I will choose the other submission for my final evaluation.", "3217861": "Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024. \n\nYou should pretend that we are now on February 29, 2024 and want to build the model that could predict future.", "3217863": "> Just get the response from the sponor. For the training or EDA or features engineer, you should only use the data before the test dataset. I think that means you should only use the dataset that covers the period from March 1, 2023, to February 29, 2024. \n> \n> You should pretend that we are now on February 29, 2024 and want to build the model that could predict future.\n\nPlease correct me if I'm wrong @drwtrading", "3217867": "Thanks! That was my understanding as we are trying to predict future. Were you using future dataset?", "3224640": "Any secrets you’re holding back? Or are you just waiting for someone else to overtake your second place while you stay stuck in the same spot?" }, "source": "meta" }