e6 user activity grows, this means new posts tend to have more activity than older ones (comments, favourites, upvotes, reposts)
At the same time, as more entries are added, older posts become harder to find and get less recognition.
I'd like to have a feature that attempts to normalize these scores that one could sort by, to reduce the effects of stat inflation on the e6 score economy.
Similarly, there is the effects of compounding scores where that which is popular is more likely to garner higher scores.
e.g. someone searching for the highest score post will find it and upvote it, furthering its ability to be found via that same method and getting more upvotes.
The snowball effect is also bad for downvote storms.
So I think having some manner of sorting by a normalized score would be very healthy for e6, maybe give a little more objectivity to the posts.

I don't know what's possible but I imagine the easiest implementation would be some factor based on post age could be done with some model of activity growth overtime.

I think this is something we really need today.

If you apply score:>2000, you can see the first was uploaded 9 years ago.
The next >2k post took 3 years to show, and was from Zonkpunch. Zonkpunch takes the next 4 >2k posts before 1 year later we get MikeInel.
Again a year later for the next 2k from Fuzzamorous.
If we look at the other end, we got one 4 days ago, 10, 16, 19, 22, 23, 24 and then finally 1 month ago.
Don't get me wrong, they're great posts and we have tons of great artists that have stuck around. No doubt if the top artists are consistently uploading frequently you'll get this same effect, but I think it's pretty clear that it's much easier to get high scores now than it used to be. I also know there's been some inferior/duplicate replacements but even so, could anyone imagine having a 5k score 5 years ago? Our first 5k score post was last year... And 6 months later we have a 10k.
With the score:>2000 search, 249 results are from the past 2 years. There are only 21 before that. Over 10X as many in 1/4th the time.

That's a pretty interesting observation, I'd never have guessed we had so much powercreep here

My main issue here is that I don't see any obvious ways of implementing a metric like that. New posts do get the bulk of attention when they're posted but if we consider only the age in the calculation then we'll just end up boosting popular old posts that have been riding the longtail for years of being in the top of the order:score and order:favcount lists

I would be very interested in seeing how this discussion goes, me studying reccomendation systems never touched on anything like this

The proposal has been made before. It's likely horrifyingly complex to implement.

What you should do is search by date range and order:score at the same time.

topic #24996

With this you can easily search your favorite tags by day, week, month, year, or arbitrary date range and keep moving along into the past by just clicking a button.

This could easily be implemented for all users.

lance_armstrong said:
The proposal has been made before. It's likely horrifyingly complex to implement.

What you should do is search by date range and order:score at the same time.

topic #24996

With this you can easily search your favorite tags by day, week, month, year, or arbitrary date range and keep moving along into the past by just clicking a button.

This could easily be implemented for all users.

The link you've provided doesn't address my concern at all.
The goal is to see ALL POSTS FROM ALL TIME PERIODS ordered in ascending/descending order following the new metric.
Not just that, but to see that new metric shown on all posts to get a better sense of their "worth", given the context of the user activity that existed at the time of their posting vs how much positive feedback it generates.
It's about adding a weighting to the scores, because statistical trends will always be. High popularity begets high popularity, poor can never get above the threshold required to stop being poor etc. etc.

The search by date range+order by score is only step 1.
You would then take the highest score of every day and then model that into some function which could then approximate the sort of user activity powercreep going on, which would be a factor to then apply in combination with votes to generate a new metric (unless you have some other user activity back-end tracking statistics)
It probably wouldn't be necessary to model that many posts, you could maybe start with the "best of" every year, and create a model with that. Go down to every month to get more precision and make adjustments, every day would probably give plenty of confidence but for "perfection" you would sample every post, as well as the vote times because if the system is implemented without vote times, and people start using this sort method, the old posts will suddenly get a huge influx of new votes, which would break the weighting completely if it's only based on post date and doesn't take into account vote date.

The weighting can of course be artificially made softer if we don't want it to have too large of an impact but I think we're sorely in need of such a feature SOON so that older posts can get the recognition they deserve and inherent merits rise above simple user growth consequences.

TIL there are 259 posts on e621 with a score of at least 2000...

...but only 39 of them are not animated...

...of which 19 were not drawn by Zackary911.

cormy1 said:
Not just that, but to see that new metric shown on all posts to get a better sense of their "worth", given the context of the user activity that existed at the time of their posting vs how much positive feedback it generates.

I know it's not the context of the thread, but I just wanna remind you and everyone else out there that your worth as an artist is never dictated by how much positive feedback you get with your art

You're always worthwhile and the world is a better place with you around

cormy1 said:
The link you've provided doesn't address my concern at all.
The goal is to see ALL POSTS FROM ALL TIME PERIODS ordered in ascending/descending order following the new metric.

I don't think your idea will be implemented. Maybe I'm wrong.

With my method, you are at least breaking up the site's history into small chunks, and then sorting those by score, so it should be more fair. And you can move through them fairly quickly.

mabit said:
... your worth as an artist is never dictated by how much positive feedback you get with your art

It is precisely because of this fact that I believe this metric is needed, as it helps level the playing field of recognition and attention when some artists get 10X as much positive feedback as what can be a work of similar worth if you weigh the effort put into it and the result of that effort, simply because they have more user activity. (this isn't really relevant from the time-based user growth factor, but the point still stands)
Of course the whole numbers game is meaningless, but it does translate to demand/income, which is something that an artist might value even when disregarding emotional impacts of point systems. It's also helpful to consumers who want to see well-received posts, but don't want to lock out half the site's content in their results because most of it falls into the tier of score that now corresponds with an average/below average post (from the perspective of their scores, because they're old and had a smaller userbase seeing them)

And the only solution is to manually browse content by time ranges?
Kind of not great, as that is a constantly expanding set of posts and limits your ability to scan them.

There's some hard pros and cons to weigh with these things, like how displaying the scores to the public at all influences their decisions on them, and you might not want that.
I'd rather the opposite, I'd rather give the public as many metrics as possible so that they can make more informed decisions. I want their decisions to be influenced, by providing more context to what they're judging.

So ultimately this is a matter of my personal preference, and how I use e6 as a consumer.
Having this metric would be a tremendously powerful filter tool for me.

wat8548 said:
TIL there are 259 posts on e621 with a score of at least 2000...

...but only 39 of them are not animated...

...of which 19 were not drawn by Zackary911.

It's funny you say this as I just recently became aware of Zackary, probably because if you filter the posts with ~female ~intersex, there are only 3 results... out of 507.
But that's a story for anever time...