I offer a small statistics and insight into this competition's preliminary leaderboard. This is not a witch hunt for any particular team or individual, this is what data scientist usually do - analyze and interpret available data.
\nI refer to my previous comment here for explanations of terms and description of categories named \"instant genius\" and \"silent companion\" and their typical characteristics.
The \"snowball\" of similarly looking preliminary private leaderboards has been rolling through various competitions since at least 9 months ago, mostly subsequently slightly trimmed by Kaggle team. Sometimes it grew significantly - as in yet not finalized NBME competition or even grew monster - in Happywhale. In this competition the snowball became an avalanche.
\nThe counts of teams/individuals showing (based on the publicly available statistical data) apparent signs of falling into at least 1 of the categories \"instant geniuses\" and/or \"silent companions\" are:
\nThere are also 5 deleted users left in the leaderboard in medal zones.
\nComparing to the previous competition here the suspiciously behaving teams became even more clustered, mainly between places 50 to 310.
\nAnother new finding is that crosschecking previous leaderboards and the 2 most recent ones it looks like there is a large pool of 1000+ accounts periodically appearing in different teams' combinations and showing signs of at least 1 of the above categories. Doesn't it look like a common backing of their activities? It reminds me the findings in the discussion here.
A week ago I asked whether I could or eventually why I shouldn't safely presume that those - here 148 - teams (sometimes repeatedly) violated Kaggle competitions' rules, for example sharing outside the team, more than 1 account per participant, circumventing Kaggle's reward system etc. Unfortunately I haven't received any response yet.
\nKaggle team also must have more details about all the teams' activity.
@maggiemd and all the other involved Kaggle team members,
\nwill the justice be finally fully served and all such malpractices will be rooted out for a better future of Kaggle competitions?
I offer a small statistics and insight into this competition's preliminary leaderboard. This is not a witch hunt for any particular team or individual, this is what data scientist usually do - analyze and interpret available data.
\nI refer to my previous comment here for explanations of terms and description of categories named \"instant genius\" and \"silent companion\" and their typical characteristics.
The \"snowball\" of similarly looking preliminary private leaderboards has been rolling through various competitions since at least 9 months ago, mostly subsequently slightly trimmed by Kaggle team. Sometimes it grew significantly - as in yet not finalized NBME competition or even grew monster - in Happywhale. In this competition the snowball became an avalanche.
\nThe counts of teams/individuals showing (based on the publicly available statistical data) apparent signs of falling into at least 1 of the categories \"instant geniuses\" and/or \"silent companions\" are:
\nThere are also 5 deleted users left in the leaderboard in medal zones.
\nComparing to the previous competition here the suspiciously behaving teams became even more clustered, mainly between places 50 to 310.
\nAnother new finding is that crosschecking previous leaderboards and the 2 most recent ones it looks like there is a large pool of 1000+ accounts periodically appearing in different teams' combinations and showing signs of at least 1 of the above categories. Doesn't it look like a common backing of their activities? It reminds me the findings in the discussion here.
A week ago I asked whether I could or eventually why I shouldn't safely presume that those - here 148 - teams (sometimes repeatedly) violated Kaggle competitions' rules, for example sharing outside the team, more than 1 account per participant, circumventing Kaggle's reward system etc. Unfortunately I haven't received any response yet.
\nKaggle team also must have more details about all the teams' activity.
@maggiemd and all the other involved Kaggle team members,
\nwill the justice be finally fully served and all such malpractices will be rooted out for a better future of Kaggle competitions?