Topic: Why so many great arts get very low upvote/favorite?

Posted under Art Talk

I've seen so many unpopular great arts in e621, they doesn't seem to get any attention like other good arts, does e621 members had unique/specific tastes for furry art?

Updated

We don't have any algorithmic discoverability beyond the daily /popular, the primary way to ensure that a post gets seen a lot after its date of upload is to make sure that it's well-tagged.

Even with over 5000 posts I'm still surprised at some things that get a good score, and some things that seem to barely get acknowledged.

Time of day it's posted can be a big factor, but sometimes I'm just not sure why the post performed that way. I guess there truly is no such thing as objectively good art.

lafcadio said:
We don't have any algorithmic discoverability beyond the daily /popular, the primary way to ensure that a post gets seen a lot after its date of upload is to make sure that it's well-tagged.

There is order:rank which has something a little more algorithmic

jonathanwolf said:
I've seen so many unpopular great arts in e621, they doesn't seem to get any attention like other good arts, does e621 members had unique/specific tastes for furry art?

I think the same... but it will be better to provide examples of that.

Most art that I like here, have very low scores.

jonathanwolf said:
I've seen so many unpopular great arts in e621, they doesn't seem to get any attention like other good arts, does e621 members had unique/specific tastes for furry art?

People only wanna look at art they like and this may mean art that belongs to a specific artist. I've experienced the opposite problem where folks would systemically downvote all posts by a single artist I uploaded and I dunno why exactly I made sure to tag the post well.

I've uploaded a few things which are technically very skilled but overwhelmingly disliked. One has -150 points.

lafcadio said:
We don't have any algorithmic discoverability beyond the daily /popular, the primary way to ensure that a post gets seen a lot after its date of upload is to make sure that it's well-tagged.

Being porn helps, as well! Not joking, very serious. This is pretty much accurate.

wolfmanfur said:
People only wanna look at art they like and this may mean art that belongs to a specific artist. I've experienced the opposite problem where folks would systemically downvote all posts by a single artist I uploaded and I dunno why exactly I made sure to tag the post well.

I think I saw that, and it's was probably hate-fans. They hate on anything by a specific artist, for example.

peacethroughpower said:
I've uploaded a few things which are technically very skilled but overwhelmingly disliked. One has -150 points.

That's actually a good sign. It shows diversity in the site's posts. Negative numbers for some content is actually a good indicator of how many people hated it enough to bother. ;)

werent dislikes meant for actual quality not "i really hate this topic"
i guess you cant really enforce that though

In addition to what has been said here, it may be influenced by simple taste. Everyone you know is different, likes different things, different drawing styles, etc. For example, I love big butts, but I hate penises. I love detailed art, though I also appreciate less detailed art. I like 3d graphics but prefer 2d more. You may like something else. And in such a site it's so cool that you never know who you'll meet and you don't know if they will have the same taste as you. Hence, such pictures have few ratings, because people simply like SPECIFIC things. Of course there are more flexible people like me, but usually people only stick to a few kinks at a time.
โ€‹

As well as all the worthwhile factors pointed out by others here, it's hard to overstate the amount of traffic that comes from off-site.

Several furry reverse image bots link to e621 as an image source, so an image that's quite popular elsewhere (e.g., on Telegram) can end up gaining a lot of traffic on e621 for that reason alone. That's just one of many ways a post can get traffic from off-site.

It's also worth noting that several animators will link to the animation's e621 post directly from FA or Twitter, because these sites -- FA and Twitter -- host animations in what the animator finds to be unacceptable quality.

Both Popular and Hot (order:rank) have the standard social-media-associated 'feature' of popular things being seen by iteratively more people because they are already popular, and unpopular things being seen by iteratively less people because they are already unpopular. People like using score and favcount as a measure of whether they should bother looking at something, as far as I can see.

I think these metrics are not wholly perverse (that is, popularity is SOMEWHAT effective at predicting quality), but they are intrinsically perverse. Goodhart's law applies. The most straightforward way to game a system in which score and favcount are treated as significant is just posting stuff that will get popular quickly. ie. be stimulating -- probably porn or particularly relatable humor, -- and be rather inoffensively mainstream, whatever that parses out to within the particular group that is e621 users. The corollary would be "if your content isn't gaming the system, it's probably losing".

This all probably is getting incrementally worse as the amount of submissions in a given time period increases.

WRT traffic driven from offsite, I guess it might be interesting to see a graph summarizing 'how someone reached a particular post' (ie. ~what domain the HTTP 'referrer' attribute mentioned) averaged over 100000 posts or so.

monroethelizard said:
As well as all the worthwhile factors pointed out by others here, it's hard to overstate the amount of traffic that comes from off-site.

Several furry reverse image bots link to e621 as an image source, so an image that's quite popular elsewhere (e.g., on Telegram) can end up gaining a lot of traffic on e621 for that reason alone. That's just one of many ways a post can get traffic from off-site.

It's also worth noting that several animators will link to the animation's e621 post directly from FA or Twitter, because these sites -- FA and Twitter -- host animations in what the animator finds to be unacceptable quality.

Is there anywhere where there's a writeup on more technical metrics and setups? I'd be quite interested in what e621 runs on.

alphamule

Privileged

TBF, I don't usually use the popular sort type thing. I mostly search by tags or from tangents when tagging/reading forums. Please don't spam forums or tag-stuff just to get views.

The real problem alluded to was basically: SNR or signal-to-noise ratio where some posts have a LOT of views and thus the votes don't tend towards fast swings, while less popular images get a vote ratio that will swing all over. I don't think we have a up/down ratio feature, though, just a sum. You can't tell 500/1000 (up versus total votes) from 4/8 or 504/1008, anyways, because they all have a ratio of 1:1 but difference of 0, but for say, 304/380 (difference of 76 and 80%) versus 204/280 (difference of 76 and ~73%) versus 280/350 (difference of 70 and 80%), it's more useful. There's also a measure using sqrt(Up^2 + down^2) to get a sort of absolute value of votes total, like the function you learned in trigonometry and linear algebra. Talking about tangents, I've seen some graph votes like an analog gauge using slope with red/green zones, haha.

peacethroughpower said:
Is there anywhere where there's a writeup on more technical metrics and setups? I'd be quite interested in what e621 runs on.

afaik, e621 was built from scratch, unlike most boorus being based off of danbooru.

magikarp said:
afaik, e621 was built from scratch, unlike most boorus being based off of danbooru.

e621 is built on Danbooru (I think there are still Danbooru references in the source code), but heavily modified over the years. One of the reasons for the 2020 site rewrite was because they wanted to move to a new version of Danbooru, but had modified the original code too much for a straight upgrade to be possible.

Some Observable Patterns I Know:

NSFW artworks on this website tend to receive more ratings than SFW ones. Consequently, if an artist focuses on creating SFW art, they may attract fewer followers.

Your profile picture features an image from a Japanese artist. The Anime style by Japanese artists is less appreciated on this platform, as mentioned in the first point. (I'm unsure how to describe the popular styles of European and American artists.)

Lastly, it's widely acknowledged that fan art is more easily discoverable than original pieces. Therefore, for an artist looking to increase their visibility, creating fan art of popular furry characters is a better choice.

peacethroughpower said:
I've uploaded a few things which are technically very skilled but overwhelmingly disliked. One has -150 points.

Is it this one ๐Ÿ˜ญ๐Ÿ˜ญ
post #4045780

foggydragon said:
Some Observable Patterns I Know:

NSFW artworks on this website tend to receive more ratings than SFW ones. Consequently, if an artist focuses on creating SFW art, they may attract fewer followers.
[โ€ฆ]

Relevant kekeflipnote https://youtu.be/nNDefI-Y3-E?si=VKPgIrqtAputWhDf

lafcadio said:
We don't have any algorithmic discoverability beyond the daily /popular, the primary way to ensure that a post gets seen a lot after its date of upload is to make sure that it's well-tagged.

I recently prototyped a very flexible custom ordering system - you could do things like order by favorites*upvotes*id/total_posts - however it is DEFINITELY not something that would be likely to be in the production version of e621 - while it does its job, it is buggy, doesn't handle unexpected inputs well, and may cause issues elsewhere.

Quite honestly, tons of technically impressive but also extremely generic and uninteresting art gets very popular here. I've also observed "thumbnail appeal" as a phenomenon too; some art is great once you blow it up from a thumbnail, but the thumbnail doesn't look too interesting on its own- and if the thumbnail doesn't look interesting, people don't look at it. Also, people tend to be far more interested in whatever animation NAME_OF_POPULAR_ANIMATOR just made even if it's just a generic loop, since posting 2D NSFW animations here is essentially free upvotes at this point. I've seen 2D art get a little tweening and instantly go from 50 upvotes for the still to 1000 in the tweened animation despite essentially being the same art.

  • 1