It doesn’t feel like the algorithm is weighted properly. I usually see the more recent comments near the top even if they have no or very few upvotes (or sometimes even in the negative), with the highly upvoted comments somewhere lower down.
I would expect Hot to feel more like Top but with extra weight given to the rate at which a comment is being upvoted, giving more visibility to new discussions in older threads.
I think comments with a high number of upvotes deserve a little more weight to avoid a 12h old comment with 60 upvotes being outcompeted by a 12m old comment with one or two upvotes. I could accept a 1h old comment with 10 upvotes being placed above a 10h old comment with 30±10 upvotes.
I would also like to be able to change the default for my account.
Everything about Lemmy; bugs, gripes, praises, and advocacy.
For discussion about the lemmy.ml instance, go to !firstname.lastname@example.org.
This is a much less extreme example, but I still feel it illustrates the point:
I don’t think a 2h old comment with no upvotes (beside the auto-upvote for new comments) should place above an 8h old comment with 4 upvotes. Whether a 4h old comment with half the upvotes of an 8h old comment should place above the latter is more debatable.
I’d like to find an example with higher numbers in a 24h window but that’s hard to come by at this time.
Edit: another example:
That’s intentional, new content should be a the top for the first few hours, otherwise no one will ever see it to vote on it. After a few days, it should average out to look like top. If you don’t do that, you get reddit’s terrible example which gives too much momentum to early rather than new comments, so every top comment is just the person to do “first!” in any post.
I got that a lot on Reedit. Every time I comment I think to myself “why bother?” New posts should be consumed as well.
But perhaps they shouldn’t be called “hot”. It maximises post data usage and therefore be called… " eco"?
Reddits sort might as well be called top, because that’s how it functions in practice, but we also have that.
hackernews and us actually factor in time to our sorting algorithm, and we both call it hot I think.