Fanboys purposely giving low rating to specific releases?

[Hirameku]

Hi fellow anime fans.. I registered today only to make this thread, I have no idea if this has been up already, if so I am truly sorry.

I have been using anidb for a very long time and love all aspects of it. ^^

But recently I noticed this and it was bothering me.

I dont know if many of you go by the ratings that people add for specific releases of an anime done by different groups, but until recently I always thought that they were pretty good for giving you an average idea of which group is best for what anime, no? Only until recently that was...

I wont mention any show but I can take this for an example.

Say that a group's score for releases of a specific show is like this. (just example)

11 people give the rating "8"
23 people give the rating "9"
8 people give the rating "10"

and then...

20 people give the rating "0"

You understand what I mean..? So I think it really should be checked by some moderator, Please? It Should be possible to see if a person has created multiple accounts to try and give many low rating, through the IP right?

So that was about it I guess, my best regards to you all.

[nwa]

Yes, it has been brought up before.

We didn't have this problem cause anidb wasn't that popular back than which means there were less tards voting and making group comments. I swear if we'd put a "No tards allowed" restriction to commenting on groups and it'd actually work in some magical way, then we'd have about 70% less comments as well. Reading the comments regularly, some of them will make baby Jesus cry.

There's no 0 vote, lowest is 1, but I guess you were only making an example.
Yes, it's possible to check for multiple accounts, we've been deleting all the votes of such multiaccount users for a long time already, but we can't monitor ALL the people 24/7 cause we have lives.. well... we don't, really, we just have other geeky stuff to do besides herding tards.

btw, I find this one to be quite funny: http://anidb.info/perl-bin/animedb.pl?s ... 53&nonav=1

[Hirameku]

Thanks for the reply, glad to see that guys are aware of it.
Many tards nowadays indeed, I mean why bother doing such a thing >< especially how it's obvious when some of the more well known quality groups get a bunch of low ratings, in the end it's all about preference I suppose.

Funny that you mention REC hehe.

http://anidb.info/perl-bin/animedb.pl?s ... 61&nonav=1

[DonGato]

I never cared about other people votes/opinions.
Unless it's a trusted and known one.
I suggest you to do the same. There is no easy fix for stupidity.
Giving a 1 to Lunar's REC release or gg's is plainly stupid.

[belgabor]

How about calculating both mean and median and marking a vote as 'maybe fishy' if they differ too much?

[Guest]

How about calculating both mean and median and marking a vote as 'maybe fishy' if they differ too much?

thats not a bad idea, using a different kind of averaging system, i do look at the groups ratings when im looking to download an anime, i dont really have time to sit and watch the same ep 4 times with 4 different groups to find out which is best!

[linktochaos]

If you're going to check the one ratings then check the ten ratings as well.

[sscilli]

This is definatly a problem, though I don't see how you can stop assholes from down voting. I would say that if you rely on the ratings to choose a group than just check the voting history and exclude the 1's, unless their are alot of lower ratings sugesting that it is a bad sub. You could do the same with the 10's to I suppose. If only everyone could vote objectivly.

[Guest]

Excluding the 1s and 10s these days doesn't leave much behind - take ergo proxy for instance: on both Ani-kraze and shinsen's vote lists around two-thirds of the votes are either 1s or 10s.
A while back I heard that thes 1 / 10 voters were going to be booted and the votes rolled back, but since then not only have the votes been removed, but things have become steadily worse. The group votes system has now become a complete joke. I've always used it to try and be impartial as possible and give useful information (eg resolution, aspect ratio etc.) because let's face it, if you're downloading say, a whole series of something, you don't want to end up with a semi-literate translition in glorious engrish and video that looks to be done by an impressionist or cubist painter when there's a really good version out there.
On top of this, I've seen enough comments in there from the fansubbers themselves to realise that at least some of them take the comments and votes here seriously, and I'd hate to think that a group might drop a series simply because of the retardedness of a group of fanboys from another, possibly sub-par, group.
I'm going to keep voting for the time being, but will probably give up (as I did with AnimeNfo's site) if some clean-up doesn't take place soon.

Meh

[Der Idiot]

A while back I heard that thes 1 / 10 voters were going to be booted and the votes rolled back

you misinterpreted our lines. we said people with MULTIPLE accoutns get their votes reset. which is what we do. if singleacount user are only voting 1/10 how the fukc should we help that? that doesn't need votereseting, but a total different approach to votecalculation, which was discussed numerous times, but with no consense on it yet.

see:
http://www.anidb.net/forum/viewtopic.php?t=3322
http://www.anidb.net/forum/viewtopic.php?t=3227
http://www.anidb.net/forum/viewtopic.php?t=2309
http://www.anidb.net/forum/viewtopic.php?t=4882

anyway wiping votes of singleaccount users is retarded. this is still anidb not nazidb. if people want to be tards with their one account then so be it.

furthermore what you expect us to do is just not doable. we are talking about

<number of groups per series> * <number of anime> devided by <number of active mods, which is less than 5>

do you see how this equation doesn't work out? once more reseting votes isn't the solution to this problem, but a more smart approach to the votecalculation.

[Guest]

Fine - if I misinterpreted what you say then ok, my bad. If single account users are only voting 1 or 10 and you can't pull them out by querying your database, then that's understandable too.

Quote: 'this is still anidb not nazidb. if people want to be tards with their one account then so be it. '

I can't see how trying to stop morons deliberately abusing the system makes it nazidb. When the voting system becomes worthless because of them, then you can proudly say 'yeah don't take any notice of any scores on here as they're worth s**t, but at least we didn't oppress anyone'.

Still, I can see it's going to be tough as 'tard' votes seem to be around 50% or more of the votes on recent titles now, and the comments are getting there too with lots of name-calling and other retardedness.

Good luck

[Der Idiot]

you know in democracy as much of a tard you are you are still allowed to vote. hence if 1 user with 1 account votes 1/10's we just don't want to stop him. there is plain and simple no reason to do so. who are we to judge the intregity of opinions of users. on a personal level i indeed hate those tards, but on a more global anidb level i just have to accept it. also because it would just take too much tiem to deal with it.

When the voting system becomes worthless because of them, then you can proudly say 'yeah don't take any notice of any scores on here as they're worth s**t, but at least we didn't oppress anyone'.

we have passed that point long ago. systems in whcih every vote is considered equal will always be abused or misused. thats why i pointed out that reseting votes is not the way to go. this goes for all votes in anidb.


anyway bickering here won't do anything. we won't reset votes of singleaccount users and we (the mods) definitely won't work ourselves through thousands of singleuseraccounts just because of messed up votes.


if you want to be productive feel free to think of some good algorythm for the votecalculation which does make a difference

[Guest]

I'm sure you've already had lots of discussion about it, but here's my 2c worth. I think that you'd probably have to utilize a combination approach including:
1. A 'user ranking' system in a similar way to Google's 'page rank' system, where an account is assigned a value according to voting pattern history and number of votes cast (updated each time a vote is cast) and the votes weighted accordingly (i.e. value drops for an extremes voter, but increasing value for number of votes cast). So for instance, a new account has very little value, but with non- extreme voting, the value increases, whereas an extreme voter's value will never rise as the number of votes increment will be cancelled by the increasing negative of the voting behavior.
2. For a particular title, the weighting would then be further adjusted if the account has issued 10/1 (i.e. opposite) on one or more another group's version of the same anime (with retrograde adjustments applied on earlier votes for other groups).
3. Lastly, the votes accumulated for each group for that amnime title would have to be weighted independently for extremes voting by a simple (or not) pattern analysis.

From what I've seen, once these algorithms are determined and implemented , they tend to be kept very secret, which is why it's actually very hard to find good examples, though I know that IMdb uses a pretty good version.

[pelican]

I'm sure you've already had lots of discussion about it

Yes.

[nstgc]

I'm sure you've already had lots of discussion about it, but here's my 2c worth. I think that you'd probably have to utilize a combination approach including:
1. A 'user ranking' system in a similar way to Google's 'page rank' system, where an account is assigned a value according to voting pattern history and number of votes cast (updated each time a vote is cast) and the votes weighted accordingly (i.e. value drops for an extremes voter, but increasing value for number of votes cast). So for instance, a new account has very little value, but with non- extreme voting, the value increases, whereas an extreme voter's value will never rise as the number of votes increment will be cancelled by the increasing negative of the voting behavior.
2. For a particular title, the weighting would then be further adjusted if the account has issued 10/1 (i.e. opposite) on one or more another group's version of the same anime (with retrograde adjustments applied on earlier votes for other groups).
3. Lastly, the votes accumulated for each group for that amnime title would have to be weighted independently for extremes voting by a simple (or not) pattern analysis.

From what I've seen, once these algorithms are determined and implemented , they tend to be kept very secret, which is why it's actually very hard to find good examples, though I know that IMdb uses a pretty good version.

new=!inaccurate nor does it mean that person is a hate-voter. The closest thing that I would suggest along the lines of (1) is weight it based on the users standard diviate in their votes. However, I have a feeling that it would be found that users with lower standard deviates tend to vote higherl. Number two would require the use of the cosine system being used to relate animes and some users tend to vote with lots of tens (which would lead to a low variance). Number three just makes it unnessisarily convoluted.

Personaly I think the system should take a fixed percetage of the highest and lowest (almost universal 10 and 1). I would also suggest taking off a little extra from the 10's as all the distrobutions I've looked at tend to have a negitive skew.

I've said this several times before, but I would also like to see a the sample mean, second moment about the mean, and third moment about the mean. This is mostly becuase it annoys me to have to do these calculations myself. Few would benifit from the extra information, but I may as well ask at the end of a related post.