{ "id": 541273, "title": "Very Low R2 score", "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/541273", "author_name": "Satya Sai Deepak Velagapudi", "post_date": "2024-10-18T13:21:20.223000", "votes": 5, "comment_count": 8, "views": 0, "content": "

The top submission as of now has r2 score of 0.0050
\nI mean that's very low
\nLooks like the target is very hard to predict
\nHow are we even sure that this is not due to randomness
\nIs it even a substantial prediction when we all are getting very low r2 score

", "messages": [ { "id": 3021417, "postDate": "2024-10-18T13:30:44.073Z", "content": "

This is only the start - we will collectively get better with time @srilakshmivelagapudi

", "rawMarkdown": "This is only the start - we will collectively get better with time @srilakshmivelagapudi ", "votes": 4, "replies": [ { "id": 3021430, "postDate": "2024-10-18T13:55:04.550Z", "content": "

oh so you are optimistic. Great

", "rawMarkdown": "oh so you are optimistic. Great" } ] }, { "id": 3021411, "postDate": "2024-10-18T13:21:20.223Z", "content": "

The top submission as of now has r2 score of 0.0050
\nI mean that's very low
\nLooks like the target is very hard to predict
\nHow are we even sure that this is not due to randomness
\nIs it even a substantial prediction when we all are getting very low r2 score

", "rawMarkdown": "The top submission as of now has r2 score of 0.0050\nI mean that's very low\nLooks like the target is very hard to predict\nHow are we even sure that this is not due to randomness\nIs it even a substantial prediction when we all are getting very low r2 score", "votes": 5 }, { "id": 3021431, "postDate": "2024-10-18T13:58:47.823Z", "content": "

To find out, you can plot the predicted time-series versus the true target time-series and have a look at how the prediction replicate the fluctuation of the target. My assumption is that the prediction has very little variance comparing to the target.

\n

I am planning to do this work, but I haven't really look into the public kernel from the top leaderboard yet.

", "rawMarkdown": "To find out, you can plot the predicted time-series versus the true target time-series and have a look at how the prediction replicate the fluctuation of the target. My assumption is that the prediction has very little variance comparing to the target. \n\nI am planning to do this work, but I haven't really look into the public kernel from the top leaderboard yet. ", "votes": 2, "replies": [ { "id": 3021475, "postDate": "2024-10-18T14:39:23.047Z", "content": "

Could you explain the reasoning behind your assumption that the prediction has significantly less variance compared to the target? I’m curious to understand the intuition.

", "rawMarkdown": "Could you explain the reasoning behind your assumption that the prediction has significantly less variance compared to the target? I’m curious to understand the intuition.", "votes": 2 } ] }, { "id": 3023782, "postDate": "2024-10-21T02:43:35.020Z", "content": "

In case of low R² values, other models like Random Forests when combined may improve performance.

", "rawMarkdown": "In case of low R² values, other models like Random Forests when combined may improve performance." }, { "id": 3022789, "postDate": "2024-10-20T00:28:19.807Z", "content": "

Isn't the test data from this competition just mock data? The hosts have this to say:

\n

\"test data: A mock test set which represents the structure of the unseen test set.\"

\n

So it could be just noise. A placeholder.

", "rawMarkdown": "Isn't the test data from this competition just mock data? The hosts have this to say:\n\n\"test data: A mock test set which represents the structure of the unseen test set.\"\n\nSo it could be just noise. A placeholder.", "replies": [ { "id": 3022894, "postDate": "2024-10-20T03:49:53.717Z", "content": "

I think that means the test file provided in dataset

", "rawMarkdown": "I think that means the test file provided in dataset\n", "votes": 1 } ] }, { "id": 3023014, "postDate": "2024-10-20T07:43:59.747Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 3021417, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2024-10-18T13:30:44.073000", "content": "

This is only the start - we will collectively get better with time @srilakshmivelagapudi

", "votes": 4, "replies": [ { "id": 3021430, "author_name": "Satya Sai Deepak Velagapudi", "author_url": "", "post_date": "2024-10-18T13:55:04.550000", "content": "

oh so you are optimistic. Great

", "votes": 0, "replies": [] } ] }, { "id": 3021431, "author_name": "SLi", "author_url": "", "post_date": "2024-10-18T13:58:47.823000", "content": "

To find out, you can plot the predicted time-series versus the true target time-series and have a look at how the prediction replicate the fluctuation of the target. My assumption is that the prediction has very little variance comparing to the target.

\n

I am planning to do this work, but I haven't really look into the public kernel from the top leaderboard yet.

", "votes": 2, "replies": [ { "id": 3021475, "author_name": "Maaax", "author_url": "", "post_date": "2024-10-18T14:39:23.047000", "content": "

Could you explain the reasoning behind your assumption that the prediction has significantly less variance compared to the target? I’m curious to understand the intuition.

", "votes": 2, "replies": [] } ] }, { "id": 3023782, "author_name": "Rookie", "author_url": "", "post_date": "2024-10-21T02:43:35.020000", "content": "

In case of low R² values, other models like Random Forests when combined may improve performance.

", "votes": 0, "replies": [] }, { "id": 3022789, "author_name": "Wilmer E. Henao", "author_url": "", "post_date": "2024-10-20T00:28:19.807000", "content": "

Isn't the test data from this competition just mock data? The hosts have this to say:

\n

\"test data: A mock test set which represents the structure of the unseen test set.\"

\n

So it could be just noise. A placeholder.

", "votes": 0, "replies": [ { "id": 3022894, "author_name": "dc260123", "author_url": "", "post_date": "2024-10-20T03:49:53.717000", "content": "

I think that means the test file provided in dataset

", "votes": 1, "replies": [] } ] }, { "id": 3023014, "author_name": "", "author_url": "", "post_date": "2024-10-20T07:43:59.747000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "3021417": "This is only the start - we will collectively get better with time @srilakshmivelagapudi ", "3021411": "The top submission as of now has r2 score of 0.0050\nI mean that's very low\nLooks like the target is very hard to predict\nHow are we even sure that this is not due to randomness\nIs it even a substantial prediction when we all are getting very low r2 score", "3021431": "To find out, you can plot the predicted time-series versus the true target time-series and have a look at how the prediction replicate the fluctuation of the target. My assumption is that the prediction has very little variance comparing to the target. \n\nI am planning to do this work, but I haven't really look into the public kernel from the top leaderboard yet. ", "3023782": "In case of low R² values, other models like Random Forests when combined may improve performance.", "3022789": "Isn't the test data from this competition just mock data? The hosts have this to say:\n\n\"test data: A mock test set which represents the structure of the unseen test set.\"\n\nSo it could be just noise. A placeholder.", "3023014": "" } }