At the time of writing this, this competition has 4,610 teams, which already makes it the 7th largest ranking-relevant competition of all time. I was curious what this actually means for the ranking points rewarded for different positions and was a bit surprised. The formula contains a multiplier x that is based on the number of teams participating.
\n$$ x = log(1 + log({Teams})) $$
\nThis means that the total number of teams has a rather small impact on the points rewarded in the competition. First I thought this is somehow unfair, because the winner of a competition with 5,000 teams has a multiplier of 0.672, while the winner in a competition with only 1,000 teams has a multiplier of 0.602. However, this doesn't seem that bad when you consider that competitions with fewer participants often have a high entry barrier and are by now means easier to win. But When thinking about teams that are not at the top of a competition, I feel like points are not distributed in a fair way. Because the last place in a competition with 1,000 teams receives only ~10% fewer points than a team that places #1,000 in a competition with 5000 teams, even though it can be argued that this is much more difficult.
\nI am not sure how this should be addressed exactly, but maybe it makes sense to work with relative percentile ranking instead of using the absolute position. On the other hand, there are many ways to get some free points by e.g. submitting public notebooks and it doesn't matter that much anyways. But still, when being able to choose, I would think a slightly different ranking approach would make sense. What do you think?
At the time of writing this, this competition has 4,610 teams, which already makes it the 7th largest ranking-relevant competition of all time. I was curious what this actually means for the ranking points rewarded for different positions and was a bit surprised. The formula contains a multiplier x that is based on the number of teams participating.
\n$$ x = log(1 + log({Teams})) $$
\nThis means that the total number of teams has a rather small impact on the points rewarded in the competition. First I thought this is somehow unfair, because the winner of a competition with 5,000 teams has a multiplier of 0.672, while the winner in a competition with only 1,000 teams has a multiplier of 0.602. However, this doesn't seem that bad when you consider that competitions with fewer participants often have a high entry barrier and are by now means easier to win. But When thinking about teams that are not at the top of a competition, I feel like points are not distributed in a fair way. Because the last place in a competition with 1,000 teams receives only ~10% fewer points than a team that places #1,000 in a competition with 5000 teams, even though it can be argued that this is much more difficult.
\nI am not sure how this should be addressed exactly, but maybe it makes sense to work with relative percentile ranking instead of using the absolute position. On the other hand, there are many ways to get some free points by e.g. submitting public notebooks and it doesn't matter that much anyways. But still, when being able to choose, I would think a slightly different ranking approach would make sense. What do you think?
How did the amount of teams change now that the merger deadline has passed?? Its now 4618, it was 4598 yesterday..
\nI agree with your post it does not scale well, but with percentile ranking if a large % of people get the same score i.e. public notebook then do you give them all the same score? In JPX right now theres like 10% of people all in bronze medal zone with same score because of same notebook.
", "rawMarkdown": "How did the amount of teams change now that the merger deadline has passed?? Its now 4618, it was 4598 yesterday..\n\nI agree with your post it does not scale well, but with percentile ranking if a large % of people get the same score i.e. public notebook then do you give them all the same score? In JPX right now theres like 10% of people all in bronze medal zone with same score because of same notebook.", "votes": -1, "replies": [ { "id": 1905330, "postDate": "2022-08-19T00:48:37.987Z", "content": "Well, I would say percentile based on rank. So #5 in a competition with 1000 teams becomes 0.005. So that would be the same as #25 in a competition with 5000 teams. I know this is probably not perfect either since #5/1000 is probably harder than #25/5000 but maybe another variable can be introduced to offset this effect.
", "rawMarkdown": "Well, I would say percentile based on rank. So #5 in a competition with 1000 teams becomes 0.005. So that would be the same as #25 in a competition with 5000 teams. I know this is probably not perfect either since #5/1000 is probably harder than #25/5000 but maybe another variable can be introduced to offset this effect." } ] }, { "id": 1905103, "postDate": "2022-08-18T18:26:09.070Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ], "comments": [ { "id": 1905243, "author_name": "JM", "author_url": "", "post_date": "2022-08-18T21:29:20.620000", "content": "How did the amount of teams change now that the merger deadline has passed?? Its now 4618, it was 4598 yesterday..
\nI agree with your post it does not scale well, but with percentile ranking if a large % of people get the same score i.e. public notebook then do you give them all the same score? In JPX right now theres like 10% of people all in bronze medal zone with same score because of same notebook.
", "votes": -1, "replies": [ { "id": 1905330, "author_name": "Fritz Cremer", "author_url": "", "post_date": "2022-08-19T00:48:37.987000", "content": "Well, I would say percentile based on rank. So #5 in a competition with 1000 teams becomes 0.005. So that would be the same as #25 in a competition with 5000 teams. I know this is probably not perfect either since #5/1000 is probably harder than #25/5000 but maybe another variable can be introduced to offset this effect.
", "votes": 0, "replies": [] } ] }, { "id": 1905103, "author_name": "", "author_url": "", "post_date": "2022-08-18T18:26:09.070000", "content": "", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "1905099": "At the time of writing this, this competition has 4,610 teams, which already makes it the 7th largest ranking-relevant competition of all time. I was curious what this actually means for the ranking points rewarded for different positions and was a bit surprised. The formula contains a multiplier *x* that is based on the number of teams participating.\n$$ x = log(1 + log({Teams})) $$\nThis means that the total number of teams has a rather small impact on the points rewarded in the competition. First I thought this is somehow unfair, because the winner of a competition with 5,000 teams has a multiplier of 0.672, while the winner in a competition with only 1,000 teams has a multiplier of 0.602. However, this doesn't seem that bad when you consider that competitions with fewer participants often have a high entry barrier and are by now means easier to win. But When thinking about teams that are not at the top of a competition, I feel like points are not distributed in a fair way. Because the last place in a competition with 1,000 teams receives only ~10% fewer points than a team that places #1,000 in a competition with 5000 teams, even though it can be argued that this is much more difficult.\nI am not sure how this should be addressed exactly, but maybe it makes sense to work with relative percentile ranking instead of using the absolute position. On the other hand, there are many ways to get some free points by e.g. submitting public notebooks and it doesn't matter that much anyways. But still, when being able to choose, I would think a slightly different ranking approach would make sense. What do you think?", "1905243": "How did the amount of teams change now that the merger deadline has passed?? Its now 4618, it was 4598 yesterday..\n\nI agree with your post it does not scale well, but with percentile ranking if a large % of people get the same score i.e. public notebook then do you give them all the same score? In JPX right now theres like 10% of people all in bronze medal zone with same score because of same notebook.", "1905103": "" } }