![\"enter](\"https://3e1cb9a1-a-62cb3a1a-s-sites.googlegroups.com/site/jonathanbouchet0928/kaggle_rank.png?attachauth=ANoY7cqWBWNVxtT5CwYl7UTwdlnnOvCkDf7X4u_sPN6-aQX0tGIaEgt7_KHW1SO8cEZfUW6lWC4XLfPJ47kCDgAY_QpU2nq8NRlfF3RyWcwaQI1f6UgavFEH6wekE5IB6x_7zCQd8wjOoHLseuFZ0QDcS9_LijnmKOT6h2YQg1sLzi2i6lagxu9kCCv7OMk4NDWPWMIPeBLYLSs-I4dogaTpIvJH9Wa0lUu7W_bMLQNSL7k3MrEu4vI%3D&attredirects=0\"){kind=icon}
Please tell me my math is wrong... Otherwise, Kaggle competition points are so broken (and therefore, Kaggle rankings too) that its very discouraging...
\n\nRSNA Pneumonia Detection Challenge
\nWhat I did? Nothing, at all, fork and submit a couple of scripts and sample submission. \nI did nothing and maybe I am going to be disqualified from this... but... thats not the point.
\nI am almost last! in this competition.
\nPoints awarded: 666
Santander Value Prediction Challenge
\nI didnt do much... but at least I put some effort, some submissions (28 actually!), active in the forums, and then... leak, but whatever, I did \"something\" and got a Bronze medal.
\nPoints awarded: 743
To make it even worst, if you look into my profile... you will see I have gotten lots of points and even silver medal with sample submissions ;). And there are people who game the system even worst XD.
\n\nHopefully someday the rankings would make sense.
", "messages": [ { "id": "413835", "postDate": "11/01/2018 16:16:49", "content": "Please tell me my math is wrong... Otherwise, Kaggle competition points are so broken (and therefore, Kaggle rankings too) that its very discouraging...
\n\nRSNA Pneumonia Detection Challenge
\nWhat I did? Nothing, at all, fork and submit a couple of scripts and sample submission. \nI did nothing and maybe I am going to be disqualified from this... but... thats not the point.
\nI am almost last! in this competition.
\nPoints awarded: 666
Santander Value Prediction Challenge
\nI didnt do much... but at least I put some effort, some submissions (28 actually!), active in the forums, and then... leak, but whatever, I did \"something\" and got a Bronze medal.
\nPoints awarded: 743
To make it even worst, if you look into my profile... you will see I have gotten lots of points and even silver medal with sample submissions ;). And there are people who game the system even worst XD.
\n\nHopefully someday the rankings would make sense.
", "rawMarkdown": "Please tell me my math is wrong... Otherwise, Kaggle competition points are so broken (and therefore, Kaggle rankings too) that its very discouraging...\n\n**RSNA Pneumonia Detection Challenge** \nWhat I did? Nothing, at all, fork and submit a couple of scripts and sample submission. \nI did nothing and maybe I am going to be disqualified from this... but... thats not the point. \nI am almost last! in this competition. \n**Points awarded: 666**\n\n[Santander Value Prediction Challenge][1] \nI didnt do much... but at least I put some effort, some submissions (28 actually!), active in the forums, and then... leak, but whatever, I did \"something\" and got a Bronze medal. \n**Points awarded: 743**\n\nTo make it even worst, if you look into my profile... you will see I have gotten lots of points and even silver medal with sample submissions ;). And there are people who game the system even worst XD.\n\nHopefully someday the rankings would make sense.\n\n\n [1]: https://www.kaggle.com/c/santander-value-prediction-challenge/", "votes": null }, { "id": "413857", "postDate": "11/01/2018 16:46:11", "content": "This is my greatest problem here. It's easy to simply fork and submit and still rise up the Kaggle hierachy. However, since I'm here to mostly learn and share what I know using the wealth of free datasets; this is not much of a hindrance although it would be fairer if we joined competitions with people actually putting in the work.
", "rawMarkdown": "This is my greatest problem here. It's easy to simply fork and submit and still rise up the Kaggle hierachy. However, since I'm here to mostly learn and share what I know using the wealth of free datasets; this is not much of a hindrance although it would be fairer if we joined competitions with people actually putting in the work.", "votes": null }, { "id": "413912", "postDate": "11/01/2018 19:12:58", "content": "Good point @NxGTR.
\n\nI have also wondered how they are going to deal with a bunch of sample submissions with zero scores on stage-2 LB. I am just waiting to see the LB finalized before I ask.
\n\nHow can I check the points awarded?
", "rawMarkdown": "Good point @NxGTR.\n\nI have also wondered how they are going to deal with a bunch of sample submissions with zero scores on stage-2 LB. I am just waiting to see the LB finalized before I ask.\n\nHow can I check the points awarded?", "votes": null }, { "id": "413924", "postDate": "11/01/2018 19:37:39", "content": "The definition of the Ranking system(if I'm not wrong) is described in details in the link below.
\n\nBasically it depends on the number of teams, teammates in your own team, and number of days elapsed since the competition deadline.
\n\nhttp://blog.kaggle.com/2015/05/13/improved-kaggle-rankings/
\n\n
You can find the math here.
", "rawMarkdown": "You can find the math [here][1].\n\n\n [1]: https://www.kaggle.com/progression", "votes": null }, { "id": "414189", "postDate": "11/02/2018 09:34:20", "content": "I think that the point decay somehow addresses this. Forking and submitting can allow you to climb the ranks, but not to a significant degree. You need that gold or silver medal to get a meaningful rank.
", "rawMarkdown": "I think that the point decay somehow addresses this. Forking and submitting can allow you to climb the ranks, but not to a significant degree. You need that gold or silver medal to get a meaningful rank.", "votes": null }, { "id": "414435", "postDate": "11/02/2018 19:06:01", "content": "One solution could be: give zero points for submissions below some quantile (50%? 25%?) on the leaderboard or which outperform some benchmark model. Needing to outperform a benchmark should help prevent sample submission or random submissions, needing to exceed some quantile could also prevent points for forking a kernel.
", "rawMarkdown": "One solution could be: give zero points for submissions below some quantile (50%? 25%?) on the leaderboard or which outperform some benchmark model. Needing to outperform a benchmark should help prevent sample submission or random submissions, needing to exceed some quantile could also prevent points for forking a kernel.", "votes": null }, { "id": "414932", "postDate": "11/03/2018 23:19:10", "content": "Perhaps counting unique solutions and giving score only to the first submit per unique solution and scores=0 to those solutions that are duplicates would solve the problem, at least partially.
", "rawMarkdown": "Perhaps counting unique solutions and giving score only to the first submit per unique solution and scores=0 to those solutions that are duplicates would solve the problem, at least partially.", "votes": null }, { "id": "414945", "postDate": "11/04/2018 00:06:32", "content": "\n\n\nI am almost last! in this competition.
\n
Darn it, you beat me again!
", "rawMarkdown": "> I am almost last! in this competition.\n\nDarn it, you beat me again!", "votes": null }, { "id": "415025", "postDate": "11/04/2018 07:29:53", "content": "At this point there is a couple of threads asking about the number used for points award. We were told it will be based on the number of teams in stage-1 but as it stands it is based on far less. Either all my basic mathematics knowledge suddenly escaped my brain or there is a need to fix this by the Kaggle team :-)
", "rawMarkdown": "At this point there is a couple of threads asking about the number used for points award. We were told it will be based on the number of teams in stage-1 but as it stands it is based on far less. Either all my basic mathematics knowledge suddenly escaped my brain or there is a need to fix this by the Kaggle team :-)", "votes": null } ], "comments": [ { "id": 413857, "author_name": "gonnel", "author_url": "", "post_date": "11/01/2018 16:46:11", "content": "This is my greatest problem here. It's easy to simply fork and submit and still rise up the Kaggle hierachy. However, since I'm here to mostly learn and share what I know using the wealth of free datasets; this is not much of a hindrance although it would be fairer if we joined competitions with people actually putting in the work.
", "votes": null, "replies": [] }, { "id": 413912, "author_name": "sheriytm", "author_url": "", "post_date": "11/01/2018 19:12:58", "content": "Good point @NxGTR.
\n\nI have also wondered how they are going to deal with a bunch of sample submissions with zero scores on stage-2 LB. I am just waiting to see the LB finalized before I ask.
\n\nHow can I check the points awarded?
", "votes": null, "replies": [ { "id": 414018, "author_name": "carloshuertas", "author_url": "", "post_date": "11/02/2018 01:25:15", "content": "You can find the math here.
", "votes": null, "replies": [] } ] }, { "id": 413924, "author_name": "jonathanbouchet", "author_url": "", "post_date": "11/01/2018 19:37:39", "content": "The definition of the Ranking system(if I'm not wrong) is described in details in the link below.
\n\nBasically it depends on the number of teams, teammates in your own team, and number of days elapsed since the competition deadline.
\n\nhttp://blog.kaggle.com/2015/05/13/improved-kaggle-rankings/
\n\n
I think that the point decay somehow addresses this. Forking and submitting can allow you to climb the ranks, but not to a significant degree. You need that gold or silver medal to get a meaningful rank.
", "votes": null, "replies": [] }, { "id": 414435, "author_name": "jlomond", "author_url": "", "post_date": "11/02/2018 19:06:01", "content": "One solution could be: give zero points for submissions below some quantile (50%? 25%?) on the leaderboard or which outperform some benchmark model. Needing to outperform a benchmark should help prevent sample submission or random submissions, needing to exceed some quantile could also prevent points for forking a kernel.
", "votes": null, "replies": [] }, { "id": 414932, "author_name": "vykhand", "author_url": "", "post_date": "11/03/2018 23:19:10", "content": "Perhaps counting unique solutions and giving score only to the first submit per unique solution and scores=0 to those solutions that are duplicates would solve the problem, at least partially.
", "votes": null, "replies": [] }, { "id": 414945, "author_name": "tunguz", "author_url": "", "post_date": "11/04/2018 00:06:32", "content": "\n\n\nI am almost last! in this competition.
\n
Darn it, you beat me again!
", "votes": null, "replies": [] }, { "id": 415025, "author_name": "sheriytm", "author_url": "", "post_date": "11/04/2018 07:29:53", "content": "At this point there is a couple of threads asking about the number used for points award. We were told it will be based on the number of teams in stage-1 but as it stands it is based on far less. Either all my basic mathematics knowledge suddenly escaped my brain or there is a need to fix this by the Kaggle team :-)
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "413835": "Please tell me my math is wrong... Otherwise, Kaggle competition points are so broken (and therefore, Kaggle rankings too) that its very discouraging...\n\n**RSNA Pneumonia Detection Challenge** \nWhat I did? Nothing, at all, fork and submit a couple of scripts and sample submission. \nI did nothing and maybe I am going to be disqualified from this... but... thats not the point. \nI am almost last! in this competition. \n**Points awarded: 666**\n\n[Santander Value Prediction Challenge][1] \nI didnt do much... but at least I put some effort, some submissions (28 actually!), active in the forums, and then... leak, but whatever, I did \"something\" and got a Bronze medal. \n**Points awarded: 743**\n\nTo make it even worst, if you look into my profile... you will see I have gotten lots of points and even silver medal with sample submissions ;). And there are people who game the system even worst XD.\n\nHopefully someday the rankings would make sense.\n\n\n [1]: https://www.kaggle.com/c/santander-value-prediction-challenge/", "413857": "This is my greatest problem here. It's easy to simply fork and submit and still rise up the Kaggle hierachy. However, since I'm here to mostly learn and share what I know using the wealth of free datasets; this is not much of a hindrance although it would be fairer if we joined competitions with people actually putting in the work.", "413912": "Good point @NxGTR.\n\nI have also wondered how they are going to deal with a bunch of sample submissions with zero scores on stage-2 LB. I am just waiting to see the LB finalized before I ask.\n\nHow can I check the points awarded?", "413924": "The definition of the Ranking system(if I'm not wrong) is described in details in the link below. \n\nBasically it depends on the number of teams, teammates in your own team, and number of days elapsed since the competition deadline.\n\nhttp://blog.kaggle.com/2015/05/13/improved-kaggle-rankings/\n\n![enter image description here][1]\n\n\n [1]: https://3e1cb9a1-a-62cb3a1a-s-sites.googlegroups.com/site/jonathanbouchet0928/kaggle_rank.png?attachauth=ANoY7cqWBWNVxtT5CwYl7UTwdlnnOvCkDf7X4u_sPN6-aQX0tGIaEgt7_KHW1SO8cEZfUW6lWC4XLfPJ47kCDgAY_QpU2nq8NRlfF3RyWcwaQI1f6UgavFEH6wekE5IB6x_7zCQd8wjOoHLseuFZ0QDcS9_LijnmKOT6h2YQg1sLzi2i6lagxu9kCCv7OMk4NDWPWMIPeBLYLSs-I4dogaTpIvJH9Wa0lUu7W_bMLQNSL7k3MrEu4vI%3D&attredirects=0", "414018": "You can find the math [here][1].\n\n\n [1]: https://www.kaggle.com/progression", "414189": "I think that the point decay somehow addresses this. Forking and submitting can allow you to climb the ranks, but not to a significant degree. You need that gold or silver medal to get a meaningful rank.", "414435": "One solution could be: give zero points for submissions below some quantile (50%? 25%?) on the leaderboard or which outperform some benchmark model. Needing to outperform a benchmark should help prevent sample submission or random submissions, needing to exceed some quantile could also prevent points for forking a kernel.", "414932": "Perhaps counting unique solutions and giving score only to the first submit per unique solution and scores=0 to those solutions that are duplicates would solve the problem, at least partially.", "414945": "> I am almost last! in this competition.\n\nDarn it, you beat me again!", "415025": "At this point there is a couple of threads asking about the number used for points award. We were told it will be based on the number of teams in stage-1 but as it stands it is based on far less. Either all my basic mathematics knowledge suddenly escaped my brain or there is a need to fix this by the Kaggle team :-)" }, "source": "meta" }