Topic: The score/favcount ratio and it's connection to an image submission's content, a study.

Posted under General

BACKGROUND
For every e621 submission users can see a value that unifies up- and downvotes by users, called "score" and the amount of users that put said submission under their favorites.
Previous studies suggest that the relation between favcount and score is dependent on how explicit the content of a submission is.
It is known that users choose submissions as their favorite if they want to see them again for whatever reason and they upvote them if they "like" the submission for whatever reason but not necessarily want to see it again in the future, or don't care if they will see it again.

OBJECTIVE
Find out if and how the relation between score and favcount of an image submission depends on how explicit the content of said image submission is.

DESIGN
We selected 100 furry related images that
are not animated,
are not funny,
are not sad,
only depict mild fetishes,
are generally not "emotional",
do not contain text,
depict at least one and at most two characters,
have a high score or high favcount. (At least one value above 50)

Those constraints in image selection are necessary to filter out factors that are known to greatly influence score and favcount but aren't well understood (yet). For example, mostly independent of their contents flash submissions show a vast increase in rating and favcount for unknown reasons.
We are aware that score is a flawed measure, since we do not know how many users downvoted a submission while favcount remains a purely positive value.
To solve this problem we developed the complex "objective perceived repulsion score compensation value" to increase our internal score calculation value.

30 randomly selected participants were shown those images in a random order and they had to rate all of them on their own "subjective eroticism scale" and "subjectively perceived repulsion scale" of 0 to 10. The values acquired by this method were unified to the "objective eroticism scale value" and the "objective perceived repulsion score compensation value":

Score (adjusted through the "objective perceived repulsion score compensation value") and favcount were written down per image and a "upvote per fav" value was calculated. This values range is above 0.1 and below 0.9 in most cases.
Then we measured how explicit an image is on a scale from 0 to 10 with the following method:

On this objective eroticism scale
0 means "fully clothed, not suggestive at all", for example:
http://e621.net/post/show/197807, this image received a 0.6 on the scale, it has a "upvote per fav" level of 0.693333
10 means hardcore fucking, for exmaple:
http://e621.net/post/show/88183, this image received a 9.6 on the scale, it has an extreme "upvote per fav" level of 0.075

RESULTS
The "upvote per fav" value is inversely proportional to the value on the subjective eroticism scale.

CONCLUSION
If an image depicts hardcore pornography, more users fav it, but don't upvote it.
If an image is completely safe, more users upvote it, but don't fav it.
Average values of the "upvote per fav" value dependent on the objective eroticism scale level were calculated with very high significance.

The full study including in depth analysis of the "objectively perceived repulsion scale", eroticism scale, image selection, calculation mechanics, math and statistical images will appear in the next Journal of Applied Behavioral Science

Updated

Your entire study, and therefor your conclusions are completely flawed.
Multiple of your factors being considered are purely subjective.
Your exclusion of less popular images fails to account for them in your "statistics".
Your sample size is absurdly small.

This post, and everything in it is completely meaningless and you have wasted your time, my time, and everyone involved's time.

Edit: Forgot to mention, you also failed to account for artistic quality, another potential factor in people's judgement on this issue.

Updated by anonymous

Hammie said:
Your entire study, and therefor your conclusions are completely flawed.
Multiple of your factors being considered are purely subjective.
Your exclusion of less popular images fails to account for them in your "statistics".
Your sample size is absurdly small.

This post, and everything in it is completely meaningless and you have wasted your time, my time, and everyone involved's time.

Edit: Forgot to mention, you also failed to account for artistic quality, another potential factor in people's judgement on this issue.

http://e621.net/forum/show/63102

Updated by anonymous

furballs_dc said:
http://e621.net/forum/show/63102

We now know why his sleep schedule is fucked up.

---

But if it is of interest, I very rarely vote anything, I fav rarely any pictures I like, I either save it or ignore it, the only things I really fav are the things I want to use later on, either to track down the artist, use as a reaction image or other reasons, fap material is always saved on one disk or another.

Updated by anonymous

Hammie said:
Your entire study, and therefor your conclusions are completely flawed.
Multiple of your factors being considered are purely subjective.
Your exclusion of less popular images fails to account for them in your "statistics".
Your sample size is absurdly small.

This post, and everything in it is completely meaningless and you have wasted your time, my time, and everyone involved's time.

Edit: Forgot to mention, you also failed to account for artistic quality, another potential factor in people's judgement on this issue.

I just wanted to know whether I should order by favcount or score when looking for masturbation material. The answer is order:favcount, as expected.

Updated by anonymous

My next semester starts this Monday, I don't have much time left to waste my time or do nonsense. This is a critical time frame.

"This is not fun times, this is sad times."

All I know about studies is from browsing pubmed.

Updated by anonymous

Ah, so you were being satirical, not literal.

Fair enough.

Updated by anonymous

Hammie said:
Ah, so you were being satirical, not literal.

Fair enough.

Somewhat of both.
I can say with complete conviction that you should order by favcount and not by score if you want to fap to popular images. Even with the countless flaws of my methodology.

Sadly, those "popularity values" by themselves are somewhat flawed, because the older an image is, the higher the score and favcount. And the higher the favcount and score, the more people find it, because they might look through other people favorites or order by those values themselves.

I enjoy statistical stuff a lot and it's my focus at work. I am writing data summarizer logic, create visual representations of statistical data and I advise management level associates on how to find the most meaningful and relevant metrics to aid them with finding the worst performing employees.
I jokingly commented that people will be fired thanks to my work, but they told me those people will receive "extensive in-company training". (That is a lie, they do get fired.)
For example, call center calls have a time between where the software recognizes a call, the moment it starts ringing at the call center agent, the moment he picks up, the moment he or the caller hangs up, the time after the call until the call center agent goes on 'ready/waiting' again. All those time frames get horribly complicated once the call center agent passes the call to another agent. We need a perspective from the caller and from the related agents.
I found one agent that absolutely sucks. His pick up time is by far the longest and his after call time is by far the longest and for some magical unknown reason he even manages to have the longest call times, making him the very worst performing employee of that call center. (call center agents of his class usually answer some yes/no questions in a very fast rotation, this guy deals with <=2 callers a minute while the absolute average is 2.8)
In the end this has nothing to do with my fake study but you, dear reader still read it and now you know a bit more about me and the horrifying things I do and that makes me happy.

Updated by anonymous

NotMeNotYou said:
But if it is of interest, I very rarely vote anything, I fav rarely any pictures I like, I either save it or ignore it, the only things I really fav are the things I want to use later on, either to track down the artist, use as a reaction image or other reasons, fap material is always saved on one disk or another.

Basically this. I almost never vote on things, since I don't see much of a point in it. So I made the number go up or down one point...who cares? Also, it takes a lot for something to go on my Favorites list. Don't understand the people who have thousands upon thousands of images favorited...

Updated by anonymous

If it's of any use/interest, the highest scored and highest favcount image I've uploaded(after removing some animated ones as per your exclusions) is this:
post #231024

Also, your method doesn't seem to exclude copyright-associated stuff, which seems to be one of the factors that affects score and favcount the most. For example, as much as I hate to pick on My Little Pony, genuinely good artwork will oftentimes get a far lower score and/or favcount than mediocre MLP artwork. >.>

Updated by anonymous

aaaaand what about the posts that had their scores nuked in the 2010 update

[edit]
I'm just kidding I dont care

Updated by anonymous

ippiki_ookami said:
aaaaand what about the posts that had their scores nuked in the 2010 update

[edit]
I'm just kidding I dont care

We'll create an additional objectively compensated value for for the 2010 score drop, add it right here, press "calculate" and here we go.
The results are now updated.

Updated by anonymous

  • 1