Sphere [he/him, they/them]

I didn’t write it; that was someone else back in the earliest days of Hexbear, back when it was still chapo.chat. I just took the SQL from the old codebase and ported it into the new one in the appropriate way to make it work.

As far as I know, it is arbitrary, though.

Your understanding is correct, yes.

Including both sorting algorithms (or a novel one in addition to the existing one) would definitely complicate the picture, though I’m not really sure how significantly. What I did was fairly simple, whereas adding a new sorting option would require changes to the UI code in addition to back-end changes like I made. It’s not impossible, but it would require more work, and would increase the odds of updates to upstream Lemmy requiring more work to merge in (if they change something in any of the parts of the code we’ve modified, there will be a merge conflict that has to be resolved).

Regarding struggle sessions, there’s no true solution, but I think having the thread fall off the feed fairly soon helps to mitigate them to at least some degree. I don’t have any data to back that up, though, so I could be wrong.

As for implementing a novel algorithm as a replacement to existing sort, I’m no better equipped than you to devise one–in fact, your graphs make clear that you’re more prepared to analyze algorithmic effects than I am–I just plugged a few values in to get a sense of what it was doing at various times after the post was made.

If you’re interested in working on something like that, it could be implemented on test.hexbear.net to see what effects it would have (if we refresh the data every so often, we can get at least some sense of how it affects post rankings). If you are, I suggest reaching out to an admin to get an invite to the dev chat. (I’m unfortunately not likely to have any time for working on Hexbear code anytime soon, though. Plus I don’t really know much of anything about Rust anyway.)

Made a few edits to the above comment after rereading your post.

I was the one who reimplemented the old Hexbear sort algorithm after we merged back to upstream Lemmy and began federating. (I did not come up with the algorithm; that was someone else. I just took it from the old codebase and modified the newer codebase to do the same thing.)

I agree with your analysis; one of my goals in making the change was to stop struggle sessions from lasting two days straight, as they had begun doing under the default Lemmy Active sort.

As for @dead@hexbear.net’s complaints, it’s true that posts need comments to survive under the new algorithm, but that is even more true in default Lemmy Active sort, which simply replaces the posted_time as used in Hot sort with the last_comment_time_necro (_necro because it stops being updated after 48 hrs, hence the cutoff where posts finally drop off the feed). So a highly-upvoted thread that only has one person commenting on it every so often will stay right at the top of the feed.

And yes, the bump comments are an effort to abuse the algorithm, it’s true. That’s pretty much exclusively done in the mutual aid comm though, and I don’t think anyone is truly opposed to that here.

It’s also worth noting that those bots do not make any difference; the bump comment itself does the bumping, and the bot comments, which happen only seconds later, do basically nothing to further boost the post. When I get around to coding a bit that will wait 30 mins to respond, I hope we can ban those other bots from the mutual aid comm, so that threads don’t get cluttered with useless garbage comments that don’t even do anything useful.

Anyway, I don’t think either algorithm will fix the problem of important posts getting lots of up votes but no comments; both algorithms will send such a post down the feed rapidly, unfortunately. So, if you think a thread is important and want it to stay on the feed, you’ve gotta make a comment on it of some sort.

Bleak. Grim. Dystopian. Nightmarish. I’m running out of adjectives for this shit, and we’re not even 3 months in…