this post was submitted on 28 Dec 2025
7 points (70.6% liked)

PieFed Meta

2098 readers
23 users here now

Discuss PieFed project direction, provide feedback, ask questions, suggest improvements, and engage in conversations related to the platform organization, policies, features, and community dynamics.

Wiki

founded 2 years ago
MODERATORS
 

I prefer platforms that hide downvotes on my content because I don’t want to see negative vote counts on my own posts for peace of mind, yet I still want to see vote totals on other people’s content. Piefed (and Lemmy) should let users hide downvotes for their own posts while keeping voting visible elsewhere.

Also, both Lemmy and Piefed lack a good outlier filter. The Top filter floods me with memes and entertaining posts, while Lemmy’s scaled feed often shows too little of actual interest. An outlier filter that detects per-community outliers would surface more relevant content without drowning feeds in content from the popular communities.

A practical implementation would maintain a rolling baseline for each community by tracking the moving average and standard deviation of post scores at a fixed time offset (for example, one hour after submission). When a new post reaches that same offset, compute its z-score by subtracting the community’s average and dividing by the standard deviation. The filter algorithm then uses this z-score as the cross-community ranking signal: instead of sorting posts by raw score, it sorts by z-score, so posts that outperform their community’s usual activity (high positive z) rise to the top regardless of absolute score. This lets the algorithm compare posts from different communities on a common scale, surfacing outliers that are unusually engaging for their community and reducing the dominance of high-traffic communities.

you are viewing a single comment's thread
view the rest of the comments
[–] PumpkinDrama@reddthat.com 1 points 1 week ago* (last edited 1 week ago) (1 children)

PieFed scaled sort does seem better. Where can I see the implementation or read how it works in plain english?

I only have around 50 posts and comments and have used it a handful of times a month for a year. I didn’t like seeing my attitude so low either; that seems like a way to drive people away if they have controversial opinions. People use votes as a like/dislike button, so deriving attitude from votes doesn’t make sense.

[–] rimu@piefed.social 9 points 1 week ago (1 children)

Scaled is just the normal 'hot' sort, plus a bit extra depending on how many subscribers the post community has compared to the largest communities (where "largest" is the average of the top 25).

Check out https://codeberg.org/rimu/pyfedi/src/branch/main/app/models.py#L812

It works better than Lemmy's IMO, but is still clunky and prone to surfacing super old content in very unpopular communities. Your weighted/outlier-based approach sounds like it would work better or at least worth trying.

People use votes as a like/dislike button

Yes and that is a pretty bad attitude, tbh. People who do that should have a low attitude score, that's how we know it's working. ;-)

[–] PumpkinDrama@reddthat.com 1 points 1 week ago* (last edited 1 week ago)

Yes and that is a pretty bad attitude, tbh. People who do that should have a low attitude score, that’s how we know it’s working. ;-)

I thought it measured votes received rather than votes given. A tooltip explaining what it means would be nice, and CSS to hide my own attitude as well.