Time for another rant. I’ve decided to start a new topic rather than necro-bump the seven-month old thread, which started off-topic anyway. Even if my proposal is rejected, I’d appreciate some feedback as to why. Thanks in advance.
The order:favcount metatag is, at present, completely and totally worthless. The reason is that when the decision was made to count only priv+ favs for order:favcount, the favcount field for existing posts was never updated. The result of this is that when I look at user:coconut order:favcount, I see post #437387 on the first page, despite it being one of the worst things I’ve ever uploaded. Even if everything I type below this is ignored, I hope at least the favcount field can be updated to a consistent value.
The real thing I want to argue for, though, is for order:favcount to count all favorites again. When I was new around here, I objected to voting being restricted to priv+ members, but in retrospect, I can see it was a good decision. However, none of the arguments that were used for putting this rule into effect apply to favorites, and in fact there are several reasons why it’s an explicitly bad idea. Here are the reasons why I think restricting order:favcount to priv+ favs is bad:
- The strong relationship between score and priv+ favcount makes it completely redundant. Score increases when a priv+ favs a picture. Not only that, but pictures uploaded before the scoring system change had their scores reset to the number of priv+ favorites. This means that, even if the favcount field were to be updated to this value, order:favcount would be almost exactly equal to order:score for older posts. This is an improvement over having crappy old posts and good new posts intermixed, sure, but it does nothing to make order:favcount more useful.
- It’s extremely counter-intuitive. The cheat sheet says that order:favcount “Orders search results in descending order based on number of favorites.” Any priv+ member that doesn’t know about how order:favcount works will look at the number of favorites and image has, note the (apparently) crazily inconsistent results immediately assume that it’s broken.
- There is no reason for users to make sockpuppets to fav pictures. This was the main catalyst behind the score system change, and it was a valid one. Score affects the Popular page, and it is displayed on the picture page itself. Which means (this being the internet) members will view it as a competition and try to get in as many votes as possible. But neither of these things is true about favorites! They don’t affect the Popular page, and non-priv+ members cannot see how many favorites an image has. Not only that, but unlike voting, faving isn’t anonymous. All of that effort and risk just to change order:favcount, a feature that most new users don’t even know exists? I cannot imagine this would become a problem.
- Order:score is limited. The first post on the second page of order:score has a score of 61. By page ten, we’re down to 39. The means that order:score groups the vast majority of Danbooru’s 65000+ images into less than 40 buckets. This is a natural consequence of the new system, and again, I agree now with the change. But consider this: since a priv+ fav increments an image’s score, that means an image’s priv+ favcount is always less than its score. Which means that, under the current system, order:favcount would be grouping into even fewer buckets, results in large swathes of images having the same favcount. For smaller groups of images, order:score is the same as order:id_desc but with a few large contiguous chunks moved around. The easy solution to this is to just allow all favs to be counted. Images with favcounts exceeding 200 are not uncommon.
- Above all, an order:favcount that counts all favorites would provide an interesting metric. Danbooru is a sea of artwork, and finding interesting posts is not easy. Order:score provides a way to probabilistically sort by quality. Order:favcount would provide a way to sort not necessarily by quality but by popularity. You may say “why would you care what non-priv+ members think?” The answer is, because any additional metric we can use to provide a different probabilistic sort increases the number of interesting images I can find.
I can see absolutely no benefit to restricting order:favcount to priv+ favs. So, is there any chance this can be changed?
Updated by EMUltra3