![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2405813%2Fb9d8cb2ba8acaf1d6f74d7891b2cab4f%2F2019-10-17%20%208.00.23.png?generation=1571295649965660&alt=media\")
I suspect 2 digits after the decimal for the LB score may not be informative enough. Only showing 2 digits makes it hard for testing new predictions, since it seems that 0.001 improvement in LB score is significant for this competition but cannot be reflected on LB.
\n\nWould it be possible to show more digits for the LB score?
\n\nPS:\nThe reason behind my suspicion is: I tried some common feature engineering approaches and it seems that the improvement introduced by engineered features is of 0.001 magnitudes instead of 0.01 magnitudes
", "messages": [ { "id": "651074", "postDate": "10/17/2019 02:47:02", "content": "I suspect 2 digits after the decimal for the LB score may not be informative enough. Only showing 2 digits makes it hard for testing new predictions, since it seems that 0.001 improvement in LB score is significant for this competition but cannot be reflected on LB.
\n\nWould it be possible to show more digits for the LB score?
\n\nPS:\nThe reason behind my suspicion is: I tried some common feature engineering approaches and it seems that the improvement introduced by engineered features is of 0.001 magnitudes instead of 0.01 magnitudes
", "rawMarkdown": "I suspect 2 digits after the decimal for the LB score may not be informative enough. Only showing 2 digits makes it hard for testing new predictions, since it seems that 0.001 improvement in LB score is significant for this competition but cannot be reflected on LB.\n\nWould it be possible to show more digits for the LB score?\n\nPS:\nThe reason behind my suspicion is: I tried some common feature engineering approaches and it seems that the improvement introduced by engineered features is of 0.001 magnitudes instead of 0.01 magnitudes", "votes": null }, { "id": "651140", "postDate": "10/17/2019 05:00:49", "content": "We generally show more decimal places after the competition has concluded. During the competition, however, we make a judgement call on the number of decimals to keep the competition exciting, as well as keeping the leaderboard informative. We'll take this into consideration should we decide to adjust the displayed precision of the LB.
", "rawMarkdown": "We generally show more decimal places after the competition has concluded. During the competition, however, we make a judgement call on the number of decimals to keep the competition exciting, as well as keeping the leaderboard informative. We'll take this into consideration should we decide to adjust the displayed precision of the LB.", "votes": null }, { "id": "651194", "postDate": "10/17/2019 07:00:55", "content": "@alex495240115 you can use small \"trick\". Go to submissions list and sort it by public lb score.
\n\n
It will show you relative difference even if lb score is the same
", "rawMarkdown": "alex495240115 you can use small \"trick\". Go to submissions list and sort it by public lb score.\n\n\n\nIt will show you relative difference even if lb score is the same", "votes": null }, { "id": "651465", "postDate": "10/17/2019 14:11:39", "content": "Thank! It's helpful!
", "rawMarkdown": "Thank! It's helpful!", "votes": null } ], "comments": [ { "id": 651140, "author_name": "addisonhoward", "author_url": "", "post_date": "10/17/2019 05:00:49", "content": "We generally show more decimal places after the competition has concluded. During the competition, however, we make a judgement call on the number of decimals to keep the competition exciting, as well as keeping the leaderboard informative. We'll take this into consideration should we decide to adjust the displayed precision of the LB.
", "votes": null, "replies": [] }, { "id": 651194, "author_name": "kyakovlev", "author_url": "", "post_date": "10/17/2019 07:00:55", "content": "@alex495240115 you can use small \"trick\". Go to submissions list and sort it by public lb score.
\n\n
It will show you relative difference even if lb score is the same
", "votes": null, "replies": [ { "id": 651465, "author_name": "alex495240115", "author_url": "", "post_date": "10/17/2019 14:11:39", "content": "Thank! It's helpful!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "651074": "I suspect 2 digits after the decimal for the LB score may not be informative enough. Only showing 2 digits makes it hard for testing new predictions, since it seems that 0.001 improvement in LB score is significant for this competition but cannot be reflected on LB.\n\nWould it be possible to show more digits for the LB score?\n\nPS:\nThe reason behind my suspicion is: I tried some common feature engineering approaches and it seems that the improvement introduced by engineered features is of 0.001 magnitudes instead of 0.01 magnitudes", "651140": "We generally show more decimal places after the competition has concluded. During the competition, however, we make a judgement call on the number of decimals to keep the competition exciting, as well as keeping the leaderboard informative. We'll take this into consideration should we decide to adjust the displayed precision of the LB.", "651194": "alex495240115 you can use small \"trick\". Go to submissions list and sort it by public lb score.\n\n\n\nIt will show you relative difference even if lb score is the same", "651465": "Thank! It's helpful!" }, "source": "meta" }