AniDB

I so pissed off by that.
I just vote 10 for Cowboy Bebop, wish i could rate 11 if there is an option for that jk.

I saw that 58 votes for 1 >_<
These people insane ?
These people must die in MISERY !!!!

To tell you the truth, im not a big fan of SciFi such as plane, robot, machine but CB is exceptional.
The storyline/plot is so freakin awesome and creative. Not to mention the action as well.
Just finish watching it recently, guess what?
It goes straight into my favorite anime list.

meis wrote:These people must die in MISERY !!!!

Killing for a vote...
Please rest in peace...

Fanboy alert >_<.

Unless somebody spamvoted the 1's, there's no problem with it having 58 1's. Every user gets 1 vote per entry and if they feel like being lame or a fanboy like yourself, then it can't be helped.

Maybe a few percent of the top and bottom votes should be cut off to get a realistic rating. Personally I don't trust people who give 10s or 1s easily...

Well personally i think its fine as it is the tards usually cancel each others aniway

Yeay! Again! Now, if only exp hadn't deleted my thread from 2004 on this exact same issue...

Rar

How the hell 10 cancels 1???

(9+1+10)/3=6.6 << 9

i have to say i am seeing the same thing..

after one semestre of statistic at university i highly believe in those "normal distributed" statistics.. because it just works!!

i feel its really a pity that a vote can be dragged down so easily only because some lame-asses vote one for an anime (and they dont vote it because its a 1 [sincerely i never voted an anime i saw a one, because it would be something i wouldnt ever finish to watch so the worse vote i ever gave was a 4 as a temporary vote]) or because they feel that way, but because they want to drag down the vote!)

i dont know how imdb works but they say to implement a method that is not only based on the calculated vote done on all votes, but

IMDb publishes weighted vote averages rather than raw data averages. Various filters are applied to the raw data in order to eliminate and reduce attempts at 'vote stuffing' by individuals more interested in changing the current rating of a movie than giving their true opinion of it.

The exact methods we use will not be disclosed. This should ensure that the policy remains effective. The result is a more accurate vote average.

(see http://www.imdb.com/ratings_explained)

i am not saying this method is working in an optimum way but what i want to say is the following:

after a certain level of votes it can be assured that the distribution of the (discreet) votes given on an anidb entry (eg a film/ova/complete series) is distributed normally ). knowing this it would be [r]relatively easy[/b] to program an algorithm who "knows" which parts of the distribution are not included in it and would filter them out or maybe give them a minor weight.


http://en.wikipedia.org/wiki/Normal_Distribution (normal distribution on wikipedia)
http://en.wikipedia.org/wiki/Binomial_distribution (binomial distribution, better for discreet values like in our case)

for a few examples please look here:
kenshin ova1 graph (center on 10, the 1s would be filtered out by the algorithm)
NGE graph (center on 10, the 1s would be filtered out by the algorithm, the other votes (2, 3, 4, etc) would fit fine in the normal distribution)

its obvious that since almost nobody has activated the fine-grained voting process, it will happen that the centre is not on one exact vote (like in this two cases were the centre is 10 and indeed there would be people who had voted more than 10, but since this is not possible they vote 10 instead and because of this really much people have voted this vote)
and it also happens that sometimes there will be no visible normal distribution at all (for example sailor moon, since there are different people (who have a different age/sex?) and not machines on which statistics can be applied on easily), but i think it would be nice if some of the perl gurus of this forum could make a check on CPAN or any other system if he finds something usuable to adjust the raw calculated value with some "adjusted" value)

i think this would be quite an efford, but if there is anybody who is willing to do this it would be really a progress in the voting system of anidb!! (the average vote will be slightly more accurate -> people will learn vote stuffing doesnt work anymore -> people will give their real vote (hopefully) -> the average vote will become mich more accurate)

just my 2 cents! im hoping some of the mods/perl gurus is reading this though

Something more easy to that I posted too. I suggested to disable the ?% of the highest and lowest votes made.(after let's say 100 votes)

http://www.anidb.net/forum/viewtopic.ph ... 6&start=15

But you also have to think that it is more normal to vote a higher vote as it should be.

well it would surely be a solution, but then it may take too much votes and distort the vote more than neccessary

As rar already said, this issue has been coming up over the last 2 years. That fact that the voting system is still like it is should tell you enough that nobody of the mods cares enough about the overall vote (the gist seems to be that nobody cares about the votes at all, because people who have watched some series will have the brains to decide on their own and don't have to rely on suspicious popularity rankings). It's highly unlikely that anything will be changed, regardless of what fancy mathematical model you propose for vote normalization.

CU,
lamer_de

about the usage of the overall vote: there are several animes switching in and out from the anidb top10, and in this case i think the overall vote is pretty important.. as an example: i discovedered the ouran host club several weeks ago only because i saw it in the top10!
about the voting model: yeah, maybe it would be too much of an efford to really implement these things, i just said to give it a try because you find many of those "fancy mathematical models" also premade for almost every used programming language, like perl in our case, and see if the results are "nicer". it was only a "hey, what do you think of this idea?" not a "your voting system is really crappy, change it ASAP!"

if rar's post was ironic then i didnt get the irony, sorry

ruxkor wrote:after a certain level of votes it can be assured that the distribution of the (discreet) votes given on an anidb entry (eg a film/ova/complete series) is distributed normally ). knowing this it would be [r]relatively easy[/b] to program an algorithm who "knows" which parts of the distribution are not included in it and would filter them out or maybe give them a minor weight.

I feel pity to the person who was teaching you statistics, hopefully it was not me Here's a hint: anime votes' distribution has a bounded support. Meaning, it's not possible for a vote to be larger than 10 or smaller than 1. And these bounds are very much binding, as the OP himself attests when he says he'd like to have voted 11. This is really as far from a normal distribution as one can get.

What matters is not the numbers, but the ranking. Spamvoting 1's and 10's in equal quantities does not affect the ranking. If we have two animes, one with a vote of 9, another with a vote of 2, and both get a couple spam votes of 1 and 10, the new vote averages will be (9+1+10)/3=6.7 and (2+1+10)/3=4.3, so they will be pulled towards the middle, but the former will always stay above the latter.

Spamvoting does create some problems of course, and some weighting could be a good thing. I doubt it would change the rankings much though, so why bother?

There is another system i once thought up (never implemented it though) give people who always vote around the average a higher weight then people who vote away from the average. However this means you need to keep tabs on who voted what, and what is that compared to the average, and how does this affect this average...

too complicated... never mind