![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F7311235%2Fb805e2672bbcb421d78e3ace1dd7033a%2Foutput.png?generation=1688128230098040&alt=media\")
Hello, everyone.
\nThis competition was my first participation in Kaggle, and this post is my first submission.
\nLuckily, I was able to go up from 43th in the public leaderboard to 20th in the private leaderboard,
\nbut the relationship between the CV scores and LB scores has greatly confused me.
In the end, I chose three submissions : ①trust my CV, ②trust public LB, ③between ① & ②
\nAs a result, I achieved the 20th position with the \"①trust my CV\" submission.
The relationship between CV score and public/private scores is as follows:
\nCV vs Public: -0.576159
\nCV vs Private: 0.506870
Although the number of features and models used were different, there seemed to be a positive correlation between the CV scores and the Private scores.
\nI am a bit disappointed that I couldn't make the best submission. Is this issue common among Kagglers?
\nIt's quite shocking, but I will learn from this experience for the next time.
It was a fun competition. Thank you.
\nRegenerate response
", "messages": [ { "id": "2324163", "postDate": "06/30/2023 12:33:44", "content": "Hello, everyone.
\nThis competition was my first participation in Kaggle, and this post is my first submission.
\nLuckily, I was able to go up from 43th in the public leaderboard to 20th in the private leaderboard,
\nbut the relationship between the CV scores and LB scores has greatly confused me.
In the end, I chose three submissions : ①trust my CV, ②trust public LB, ③between ① & ②
\nAs a result, I achieved the 20th position with the \"①trust my CV\" submission.
The relationship between CV score and public/private scores is as follows:
\nCV vs Public: -0.576159
\nCV vs Private: 0.506870
Although the number of features and models used were different, there seemed to be a positive correlation between the CV scores and the Private scores.
\nI am a bit disappointed that I couldn't make the best submission. Is this issue common among Kagglers?
\nIt's quite shocking, but I will learn from this experience for the next time.
It was a fun competition. Thank you.
\nRegenerate response
", "rawMarkdown": "Hello, everyone.\n\nThis competition was my first participation in Kaggle, and this post is my first submission.\n\nLuckily, I was able to go up from 43th in the public leaderboard to 20th in the private leaderboard, \nbut the relationship between the CV scores and LB scores has greatly confused me.\n\nIn the end, I chose three submissions : ①trust my CV, ②trust public LB, ③between ① & ②\nAs a result, I achieved the 20th position with the \"①trust my CV\" submission.\n\nThe relationship between CV score and public/private scores is as follows:\n**CV vs Public: -0.576159**\n**CV vs Private: 0.506870**\n\n\n\nAlthough the number of features and models used were different, there seemed to be a positive correlation between the CV scores and the Private scores.\n\nI am a bit disappointed that I couldn't make the best submission. Is this issue common among Kagglers?\nIt's quite shocking, but I will learn from this experience for the next time.\n\nIt was a fun competition. Thank you.\n\n\n\n\n\n\nRegenerate response", "votes": null }, { "id": "2324543", "postDate": "06/30/2023 17:08:13", "content": "Reading through the solutions, I believe most teams, including many top-ranks, fail to select their highest-private-score submissions. The most robust case is the 1st position - 0.705 for all CV, public and private LB. What a beast!
", "rawMarkdown": "Reading through the solutions, I believe most teams, including many top-ranks, fail to select their highest-private-score submissions. The most robust case is the 1st position - 0.705 for all CV, public and private LB. What a beast!", "votes": null }, { "id": "2324976", "postDate": "07/01/2023 04:05:19", "content": "it would be interesting to have an indicative LB based on the \"Highest-Private-Scores\". For each team, the best private score would be selected and a ranking would be published. We all know our own best Private Score but we don't see the other participants best private scores.
", "rawMarkdown": "it would be interesting to have an indicative LB based on the \"Highest-Private-Scores\". For each team, the best private score would be selected and a ranking would be published. We all know our own best Private Score but we don't see the other participants best private scores.", "votes": null } ], "comments": [ { "id": 2324543, "author_name": "hoangnguyen719", "author_url": "", "post_date": "06/30/2023 17:08:13", "content": "Reading through the solutions, I believe most teams, including many top-ranks, fail to select their highest-private-score submissions. The most robust case is the 1st position - 0.705 for all CV, public and private LB. What a beast!
", "votes": null, "replies": [ { "id": 2324976, "author_name": "gehallak", "author_url": "", "post_date": "07/01/2023 04:05:19", "content": "it would be interesting to have an indicative LB based on the \"Highest-Private-Scores\". For each team, the best private score would be selected and a ranking would be published. We all know our own best Private Score but we don't see the other participants best private scores.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2324163": "Hello, everyone.\n\nThis competition was my first participation in Kaggle, and this post is my first submission.\n\nLuckily, I was able to go up from 43th in the public leaderboard to 20th in the private leaderboard, \nbut the relationship between the CV scores and LB scores has greatly confused me.\n\nIn the end, I chose three submissions : ①trust my CV, ②trust public LB, ③between ① & ②\nAs a result, I achieved the 20th position with the \"①trust my CV\" submission.\n\nThe relationship between CV score and public/private scores is as follows:\n**CV vs Public: -0.576159**\n**CV vs Private: 0.506870**\n\n\n\nAlthough the number of features and models used were different, there seemed to be a positive correlation between the CV scores and the Private scores.\n\nI am a bit disappointed that I couldn't make the best submission. Is this issue common among Kagglers?\nIt's quite shocking, but I will learn from this experience for the next time.\n\nIt was a fun competition. Thank you.\n\n\n\n\n\n\nRegenerate response", "2324543": "Reading through the solutions, I believe most teams, including many top-ranks, fail to select their highest-private-score submissions. The most robust case is the 1st position - 0.705 for all CV, public and private LB. What a beast!", "2324976": "it would be interesting to have an indicative LB based on the \"Highest-Private-Scores\". For each team, the best private score would be selected and a ranking would be published. We all know our own best Private Score but we don't see the other participants best private scores." }, "source": "meta" }