I made 8 submissions and I could go better than 0.27.
", "messages": [ { "id": 748267, "postDate": "2020-02-17T10:33:33.383Z", "content": "The test result is based on past NCAA result. Theoretically, you can dig back all the past result and get a better result. It will be very hard and the model may not perform well in next stage. Or those team that got 0.0 are previous winner.
", "rawMarkdown": "The test result is based on past NCAA result. Theoretically, you can dig back all the past result and get a better result. It will be very hard and the model may not perform well in next stage. Or those team that got 0.0 are previous winner. ", "votes": 1, "replies": [ { "id": 748272, "postDate": "2020-02-17T10:39:27.120Z", "content": "oh thanks @dexter6855 I almost lost my mind over this. Some of them have 0.0 with only 1 submission and that had me startled :o
", "rawMarkdown": "oh thanks @dexter6855 I almost lost my mind over this. Some of them have 0.0 with only 1 submission and that had me startled :o" }, { "id": 748282, "postDate": "2020-02-17T10:50:38.490Z", "content": "If you see any team with 100% accuracy in any competition, it is safe to assume that they some how got the entire test data and overfit their model. It happened more than once. It is very annoying.
", "rawMarkdown": "If you see any team with 100% accuracy in any competition, it is safe to assume that they some how got the entire test data and overfit their model. It happened more than once. It is very annoying. ", "votes": 1 }, { "id": 748285, "postDate": "2020-02-17T10:53:42.927Z", "content": "Hmm.. I was aware of something acrid but I didnt know that we can refer last year's data to improve scores. Thanks for the heads up :)
", "rawMarkdown": "Hmm.. I was aware of something acrid but I didnt know that we can refer last year's data to improve scores. Thanks for the heads up :)" }, { "id": 748303, "postDate": "2020-02-17T11:13:18.020Z", "content": "Seems like it is possible to get very close to 0.0 without overfit the model. I'm pretty surprise it works.
", "rawMarkdown": "Seems like it is possible to get very close to 0.0 without overfit the model. I'm pretty surprise it works.", "votes": 1 }, { "id": 748310, "postDate": "2020-02-17T11:17:52.470Z", "content": "It does but the actual results will be calculated based on the final results of the NCAA competition when it ends. So these results may be absolutely futile
", "rawMarkdown": "It does but the actual results will be calculated based on the final results of the NCAA competition when it ends. So these results may be absolutely futile" } ] }, { "id": 748229, "postDate": "2020-02-17T09:47:02.120Z", "content": "I made 8 submissions and I could go better than 0.27.
", "rawMarkdown": "I made 8 submissions and I could go better than 0.27.", "votes": 1 }, { "id": 748584, "postDate": "2020-02-17T17:37:47.190Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 748267, "author_name": "dexter6855", "author_url": "", "post_date": "2020-02-17T10:33:33.383000", "content": "The test result is based on past NCAA result. Theoretically, you can dig back all the past result and get a better result. It will be very hard and the model may not perform well in next stage. Or those team that got 0.0 are previous winner.
", "votes": 1, "replies": [ { "id": 748272, "author_name": "Vaibhav Saxena", "author_url": "", "post_date": "2020-02-17T10:39:27.120000", "content": "oh thanks @dexter6855 I almost lost my mind over this. Some of them have 0.0 with only 1 submission and that had me startled :o
", "votes": 0, "replies": [] }, { "id": 748282, "author_name": "dexter6855", "author_url": "", "post_date": "2020-02-17T10:50:38.490000", "content": "If you see any team with 100% accuracy in any competition, it is safe to assume that they some how got the entire test data and overfit their model. It happened more than once. It is very annoying.
", "votes": 1, "replies": [] }, { "id": 748285, "author_name": "Vaibhav Saxena", "author_url": "", "post_date": "2020-02-17T10:53:42.927000", "content": "Hmm.. I was aware of something acrid but I didnt know that we can refer last year's data to improve scores. Thanks for the heads up :)
", "votes": 0, "replies": [] }, { "id": 748303, "author_name": "dexter6855", "author_url": "", "post_date": "2020-02-17T11:13:18.020000", "content": "Seems like it is possible to get very close to 0.0 without overfit the model. I'm pretty surprise it works.
", "votes": 1, "replies": [] }, { "id": 748310, "author_name": "Vaibhav Saxena", "author_url": "", "post_date": "2020-02-17T11:17:52.470000", "content": "It does but the actual results will be calculated based on the final results of the NCAA competition when it ends. So these results may be absolutely futile
", "votes": 0, "replies": [] } ] }, { "id": 748584, "author_name": "", "author_url": "", "post_date": "2020-02-17T17:37:47.190000", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "748267": "The test result is based on past NCAA result. Theoretically, you can dig back all the past result and get a better result. It will be very hard and the model may not perform well in next stage. Or those team that got 0.0 are previous winner. ", "748229": "I made 8 submissions and I could go better than 0.27.", "748584": "" } }