Vote normalization

[Elias]

[*]User must have no more than 10% '10' votes


Users like this would pass the first two rules.

and btw i would pass all 4 rules, changed alot lol

Nope. Rule 3 not passed, you have 17 votes of 10, more than 10% of 157 total votes.
You will be filtered if those rules will be applied.

[Skywalka]

I also think it is ok to find out who has a pretty strange voting pattern, and honestly I feel EXP should make up his mind what he thinks is strange.

And I feel the discussion is getting pretty old. Every three weeks there is a new thread about this and while you all know that EXP is the only person who can do something about this, nobody seems to really care to directly appreoach him. The most I've seen was "EXP hasn't said anything about it" and that wa in this thread here. In fact he never said anything about this yer if I remember correctly which is pretty funny considering the fact that there was so much discussion about this already.

You folks will end up discussing this forever if you don't get your priorities right

Those who don't even discuss it anymore have made up their minds and are tired of re-stating them for every new user dropping in there and yelling for somebody to "§)/"§ remove 1 votes for Naturd0!!!!!1!11!!" *snicker*

[OnegaiNL]

[*]User must have no more than 10% '10' votes


Users like this would pass the first two rules.

and btw i would pass all 4 rules, changed alot lol

Nope. Rule 3 not passed, you have 17 votes of 10, more than 10% of 157 total votes.
You will be filtered if those rules will be applied.

hmm maybe your right... but still 10 % is a bit low ... b/c those were all the legendary anime i've seen =/ and besides i don't really call Onegai Twins OVA and Teacher OVA a real anime.... it's kinda part of those 2 series so actually it's 15 ... that would mean i'm at like 10.5 %

20 % should be efficient i think.

[Elandal]

I think the four rules listed should be mostly enough to handle the bad votes (bad as in "the person either didn't read the vote descriptions, decided to not honor them, or just hasn't watched enough to make even an educated guess"). I'd set the numbers pretty low, maybe 10 votes and 10 complete animes in mylist.
I would add a fifth rule just in case, though: no value has more than one third of the votes. That is, giving five sevens, then one for each of 10, 1, 9, 5, and 3 would also disqualify. I don't think someone who thinks half the anime is of exactly same quality should really be considered to have a viable opinion..

My votes come to about:
10 6%
9 18%
8 25%
7 19%
6 20%
5 8%
4 2%
3 1%
and one 1.
I think my vote distribution is a little too much on the high end, but also I don't give much low votes because if I truly hate something, I don't watch it to the end and thus can't (IMO) vote for it.. I feel that it's unfair to vote on something I haven't watched completely - the beginning of a series may not tell enough of the series to properly appraise it.

[Elandal]

@OnegaiNL

If you've read the vote descriptions, you should understand what a 10 is. 20% @10 is just too much.
My problem with tens is that I'm not sure which of the 7 ghiblies to drop to 9. They're quite different, so it's hard to pin them against eachother..

Another thing is that rating movies and series happens on completely different ways. It might be useful to consider series and movies in different categories, giving us both "top10 movies" and "top10 series" listings. This would of course introduce problems with OVAs, but they could probably be split so that single-episode ovas would go to movie listing, multi-episode ovas to series listing. Even if a four ep ova is in length close to that of a movie, it has to work differently as the audience is expected to not watch them in a single session (as with movie) but in multiple sessions, with time between them.

[OnegaiNL]

@OnegaiNL

If you've read the vote descriptions, you should understand what a 10 is. 20% @10 is just too much.
My problem with tens is that I'm not sure which of the 7 ghiblies to drop to 9. They're quite different, so it's hard to pin them against eachother..

Another thing is that rating movies and series happens on completely different ways. It might be useful to consider series and movies in different categories, giving us both "top10 movies" and "top10 series" listings. This would of course introduce problems with OVAs, but they could probably be split so that single-episode ovas would go to movie listing, multi-episode ovas to series listing. Even if a four ep ova is in length close to that of a movie, it has to work differently as the audience is expected to not watch them in a single session (as with movie) but in multiple sessions, with time between them.

well the problem with this is ... that would mean if you hhave only seen 10 anime series and you vote 1 a 10 that you would be filtered out... that's just totally crap, and besides if you're not able to vote for the coolest animes you've seen from particular genre's than it's kinda messed up isn't it?

[Elandal]

Not being counted for top10 or anime hint lists doesn't mean you can't vote what you want to.
Also, if you've only seen a few animes and they happen to be the very best of the best, I'd really wait until you have seen a little more before considering your votes valid.

Now, do you remember what the vote descriptions said? Start by voting 3, 5, 7, and 9. When you've seen more, you'll know whether something you gave a 9 in the beginning is actually worth a 10 instead.

Now, I have seen very few Korean movies. One of them was really good. However, if I don't watch more Korean movies, can I really say that the movie was the best thing since sliced bread? Maybe there are lots of good Korean movies and the one I saw first was just an indication of the general quality, and should be given maybe an 8 instead of 10 when comparing it to other Korean movies (as opposed to comparing it to general Hollywood movies)?

Similarly, if you've seen animated films and series from US, and the first few animes you see happen to be better than most animes (maybe 8 or even 9, or maybe you were shown something truly worth a 10), you will probably vote them compared to US animated movies. Assuming you consider US animated movies mostly lacking in character development and story, you'd easily give half the anime out there a very high score. But when you've seen ten animes, you might notice how different they are from your average disney flick, and might be able to compare them to eachother and instead of giving all of them scores 8-10 would rather give the average disney flick a 4-6 instead of 7? If you can't compare anime with other anime after ten titles, hopefully you can do that after 20. But at some point you should be able to do it and would have been adjusting your votes accordingly, thus becoming eligible for eg. anime top10 vote counting and so on.
If you never manage to sort our your votes so that they're meaningful for others to consider, then they shouldn't be used for those purposes. They may be meaningful to you still, just not others. In that case, you'll have your votes just as you wanted them, and others aren't bothered by having your abnormal voting habits weighing in on the lists they use for choosing which great animes they should start watching next.

Note that anime hints aren't very useful to you either if your votes aren't reasonably distributed. But that's your problem, not a general problem.


EDIT:
Oh yes, did you notice what it says on anime hints -page? "You should have voted for at least 50 animes before using this feature!" - to me this indicates that before your votes can be used for statistical analysis, a lot of meaningful votes are needed. And, anime hints and top10 lists and so on are just that: statistical analysis. I was trying to be lenient with the minimum of 10 votes and 10 animes in mylist, hoping that some people could've forged meaningful generally useful opinions already after ten titles.

[DonGato]

I think what you (Elandal) wants to take care off would be fixed by the minimum anime in list filter. We don't need to also take care of so restricted percentages for votes (like only 10% for 10s, even if I would never reach it).

EDIT: just read your edit so I think you already agree with this.

[Elandal]

We definitely do need to take into account the vote distribution. Otherwise lists like OnegaiNL's old list (a hundred and something of 10s only) would qualify while it definitely should not. That kind of vote distribution is absurd and not meaningful to the least.
Also, that filter wouldn't even affect most who would otherwise qualify - those who have come to meaningful conclusions about the qualities of a lot of anime are very unlikely to think that a lot of the anime available really has changed their understanding of anime, life, self, the world, and so on.

[OnegaiNL]

We definitely do need to take into account the vote distribution. Otherwise lists like OnegaiNL's old list (a hundred and something of 10s only) would qualify while it definitely should not. That kind of vote distribution is absurd and not meaningful to the least.
Also, that filter wouldn't even affect most who would otherwise qualify - those who have come to meaningful conclusions about the qualities of a lot of anime are very unlikely to think that a lot of the anime available really has changed their understanding of anime, life, self, the world, and so on.

yea my old list was kinda ridicilous lol ... but i was bored so i changed alot... but still 17 of 157 votes = 10.... because those were really the best i've seen from those particular genre's

and if i think it's the best it should deserve a 10 until some other new anime proves it's better.

if you only eat 1 kind of cheese you will think it's the best there is... because you haven't tasted the rest yet.

[Elandal]

Welll.. I still maintain that if you indeed have >10% 10s, your voting habits aren't suitable for use in statistical analysis without some form of normalization, or you haven't seen enough to judge the titles in meaningful way.

Really, I don't even expect bell curves here. Just that you won't give more to the very highest or very lowest than one per number of discrete vote options. I find that very reasonable.
Also, I believe adding the max 1/3rd (or maybe 3/n where n is the number of discrete vote options, which would actually yield 30% instead in this case; this just in case the number of vote options would change in the future) per vote option would be very reasonable.

Note that even with these rules in place valid votes would be considered invalid and discarded by some market research answer analysis methods I've implemented (not designed - they were designed by PhD's and I just wrote programs that implemented the algorithms).

Now, the votes may be useful to YOU. That does NOT mean that they can be used in any meaningful way for analysis within a population.

[OnegaiNL]

Welll.. I still maintain that if you indeed have >10% 10s, your voting habits aren't suitable for use in statistical analysis without some form of normalization, or you haven't seen enough to judge the titles in meaningful way.

Really, I don't even expect bell curves here. Just that you won't give more to the very highest or very lowest than one per number of discrete vote options. I find that very reasonable.
Also, I believe adding the max 1/3rd (or maybe 3/n where n is the number of discrete vote options, which would actually yield 30% instead in this case; this just in case the number of vote options would change in the future) per vote option would be very reasonable.

Note that even with these rules in place valid votes would be considered invalid and discarded by some market research answer analysis methods I've implemented (not designed - they were designed by PhD's and I just wrote programs that implemented the algorithms).

Now, the votes may be useful to YOU. That does NOT mean that they can be used in any meaningful way for analysis within a population.

why isn't it suitable? should i just even lower it more?
best anime i have deserves a 6 or something? come on... if it's like that just f*cking remove the 10, not useful for statistics yea right... have you even read what i wrote?

10 % is too low..... i'm not going to change my votes again, they're right now and not planning to change them just b/c some f*cks think that it's not useful for statistics blabla

[DonGato]

I think Elandal is not taking in account an alternate pattern used by some people here that rate anime by genre. So it's logical them having more 10s than us that rate all anime instead of by genre. That's why I said we should be more lax with the limits. If someone watched 150 series and think 20 deserve a 10 I don't think I have the right to complain to him. Different would be if he voted 75 or any other high number. That's my opinion on the matter.

Anyway, I'm already glad to have a filter for 1/10 voters and ghost accounts. Having also a way of detecting weird deviations would be a nice thing also (like I said before, an anime with 500 votes, 480 above 8 and the rest bellow 2).

[kidan]

I agree with the idea of applying restrictions based on amount and distribution of votes.

But I'd like to see another feature added as well: 0.5-ratings for ratings > 5.

Why this?
- Most of our voting dynamics range from 5 to 10, as there are not that many really bad anime. So right now we limit our accuracy to only about 6 of 10 values. With the additional votes of 5.5, 6.5, 7.5, 8.5, 9.5 we'd have 11 likely-to-vote steps.
- I've experienced quite a few anime which would not be a real 8, but giving them a 7 would not be fair as well. A 7.5 would be the best then.
- Or If there's a anime which really changed your thinking, but it had minor flaws it could get a 9.5.

Why not for ratings < 5 ?
- there are not that many anime cluttering up this region
- no need to distinguish one crap from another to such an extend (imho).
- did any of you miss a 3.5 fo far? I didn't.

[CheesyP18]

so why hasnt exp given an opinion yet?

Maybe we should ask him the next time someone sees him in IRC, its pretty rare though