I love the graphs. I love the enthusiasm you have.
I see a problem with your methodology. You are comparing the algorithm on posts which start with 1000 upvotes and never changes. Posts start with 1 upvote and the number of upvotes grows based on visibility in the algorithm. The algorithm used influences the number of upvotes that a post will get.
A post with 0 comments can not reach 1000 upvotes. Comparing the decay of 2 posts with the same number of upvotes does not make sense because the algorithm also changes the upvotes that a post will get. The lack of comments causes lack of visibility which causes diminishing upvote rate, which all compounds. A faster decay means that a post with no comments loses visibility even sooner.
From my experience, posts are primarily viewed in 2 ways, on ‘new’ sort and on ‘active’ sort. The front page of ‘new’ sort lasts around 1 hour typically and then users won’t upvote the post any more from ‘new’ sort. After 1 hour, the post only gains upvotes if it is visible in the ‘active’ algorithm. A post with 0 comments never even sees the front page of ‘active’ sort, therefore stops gaining upvotes at all after 1 hour. To compare the algorithms accurately, you would need to have a new function where a post only gains upvotes when the post is visible on the algorithm front page after the first hour.
I think it’s ineffectual to try to cull controversial threads using an algorithm. This purpose was originally served by downvotes. This function can likely only be replaced by active moderators, ie locking posts with excessive arguments.
The decay being exponential seems unnecessarily aggressive. Why can’t the decay be linear and still reach a value of 0 after 24 hours?
example: SCORE * ((24 - X) / 24), where X is the age of the post in hours.
Editing my post because I want to be precise and I want to be respectful, but some of my thoughts feel abstract. My complaint is that posts with important news are dying after 1-2 hours. My speculation is that exponential decay increases the chances that important news stories are disappeared into the void within 1-2 hours. Whether hexbear’s algorithm is giving a lower hypothetical score to controversial posts is a tangent.
I love the graphs. I love the enthusiasm you have.
I see a problem with your methodology. You are comparing the algorithm on posts which start with 1000 upvotes and never changes. Posts start with 1 upvote and the number of upvotes grows based on visibility in the algorithm. The algorithm used influences the number of upvotes that a post will get.
A post with 0 comments can not reach 1000 upvotes. Comparing the decay of 2 posts with the same number of upvotes does not make sense because the algorithm also changes the upvotes that a post will get. The lack of comments causes lack of visibility which causes diminishing upvote rate, which all compounds. A faster decay means that a post with no comments loses visibility even sooner.
From my experience, posts are primarily viewed in 2 ways, on ‘new’ sort and on ‘active’ sort. The front page of ‘new’ sort lasts around 1 hour typically and then users won’t upvote the post any more from ‘new’ sort. After 1 hour, the post only gains upvotes if it is visible in the ‘active’ algorithm. A post with 0 comments never even sees the front page of ‘active’ sort, therefore stops gaining upvotes at all after 1 hour. To compare the algorithms accurately, you would need to have a new function where a post only gains upvotes when the post is visible on the algorithm front page after the first hour.
I think it’s ineffectual to try to cull controversial threads using an algorithm. This purpose was originally served by downvotes. This function can likely only be replaced by active moderators, ie locking posts with excessive arguments.
The decay being exponential seems unnecessarily aggressive. Why can’t the decay be linear and still reach a value of 0 after 24 hours?
example: SCORE * ((24 - X) / 24), where X is the age of the post in hours.
Editing my post because I want to be precise and I want to be respectful, but some of my thoughts feel abstract. My complaint is that posts with important news are dying after 1-2 hours. My speculation is that exponential decay increases the chances that important news stories are disappeared into the void within 1-2 hours. Whether hexbear’s algorithm is giving a lower hypothetical score to controversial posts is a tangent.