New rating calculations/system

Post by **pelican** » Thu Jan 11, 2007 7:01 pm

The system's only as random as the mind from which it originated, I assure you. However there is NO system that is immune to ballot stuffing except for literally fixing the ratings so we want to keep as great an edge over these people (and their hordes of little "brothers") as possible.

Also, this is only a first draft. Expect more changes later.

(And yes, it applies to all ratings. Anime (permanent and temporary), episode, group, group for anime, and, indeed, reviews.)

ngc109 · Post by **ngc109** » Thu Jan 11, 2007 8:47 pm

A quick question about the ratings: Why is it that when I view the info page on an anime (lets take Inukami for an example), it shows the rating as 7.56, yet when I click on the graph of votes, it says they average out to 8.66? Basically the rating that it shows under the graphs is completely different from what is shown on the animelist.

Post by **pelican** » Thu Jan 11, 2007 8:48 pm

Because one says "average" and the other says "rating". Nowhere does it say that "rating" means "average". (If it does, please tell us so that that can be corrected.) Also, like everything else on the vote graphs, the average is always up to date, unlike the ratings.

ngc109 · Post by **ngc109** » Thu Jan 11, 2007 9:10 pm

Ah, thats what I get for having my wishlist as the default bookmark instead of the main page. The new announcement says the old ratings are still available via the vote graphs...thats why they were different. My fault for not reading

nstgc · Post by **nstgc** » Fri Jan 12, 2007 2:04 am

This question is a bit different then the previous questions. I simply wish to know if the new system takes into account all users and all their animes as opposed to only those votes cast for the anime in question. This shouldn't give ballot suffers any advantage. If anything it should tell them "your screwed". If my assumption that the folks at AniDB did a half respectible job, then we won't be able to figure out the system just by looking at the data.

Post by **pelican** » Fri Jan 12, 2007 2:23 am

Sorry, I'm really not giving out any information. Any datum about this system is a weapon in the wrong hands and there's no avoiding that other than by completely rigging it.

laxxy · Post by **laxxy** » Fri Jan 12, 2007 2:44 am

Surely no system is immune, regardless of whether it is known or not.
still it should be possible to design a system that while being open would require a good deal of (probably easily detectable) effort to stuff the ballots significantly. I can think of a few possibilities related to weighting and normalizing users' votes (ie considering each user's voting patterns in determining the effect of his vote), perhaps that is what is being done already...

nstgc · Post by **nstgc** » Fri Jan 12, 2007 2:47 am

While I won't try and get anything out you, I will say that your behavior could lead some to believe that the current/new system is not quite loser proof (meaning gaping wound type flaw) which could encourage the wrong people, and discourage those such as myself. I'll just have to answer this question myself, when ever I get time. As I said, it shouldn't be too difficult to find out since I'm not actually looking for the algorithms. Regardless of any of that, the old system didn't work thanks to certain morons so even if this one is "idiot resistant" its an improvement. Even I'll give you that and I extremely hard on others when it comes to mathematical reasoning.

In short, I think the system is a mathematical mess still, but you guys are working on it and it is response to a problem. This is neither negitive nor positive feed back, so don't take it either way.

Website · Post by **Rar** » Fri Jan 12, 2007 2:59 am

Peeps, you're missing the point that any voluntary poll on the internet with no identity verification involved is about as valid as reading tea leaves. Our aim is to keep the numbers somewhat useful, not magically make them mean something.

Rar

laxxy · Post by **laxxy** » Fri Jan 12, 2007 4:53 am

hey, all numbers mean *something*. Even knowing that anime A has a larger number of rabid fanboys compared to anime B is useful

Post by **PetriW** » Fri Jan 12, 2007 7:45 am

laxxy wrote:hey, all numbers mean *something*. Even knowing that anime A has a larger number of rabid fanboys compared to anime B is useful

Even when it's a larger number of ONE fanboy?

laxxy · Post by **laxxy** » Fri Jan 12, 2007 8:36 am

PetriW wrote:
laxxy wrote:hey, all numbers mean *something*. Even knowing that anime A has a larger number of rabid fanboys compared to anime B is useful
Even when it's a larger number of ONE fanboy?

Perhaps what counts here is an integral coefficient: (Number)*(Rabidness).
If an anime attracts one rabid enough to register dozens of accounts to increase its' rating by a couple decimal digits, maybe there's something about it

:)
A system that would naturally underweight voters with very volatile opinions and low number of votes would make the impact of dummies much smaller, so the rabidness value required to affect the ratings would be really high

axeps2 · Post by **axeps2** » Sat Jan 13, 2007 3:06 am

Nice to see attempts at improving the system..
Continue with the great work..

And people really had the time to register lots of accounts just to change votings...? wow. I envy them for their free time.I still had tons of anime unwatched..

But... KAA ,only a 7.99 group...

*Bang wall*

I know you guys do not want to leak the workings of the new system but is it possible to say whether 1s and 10s are totally ignored or merely given lesser weightings..?

Post by **pelican** » Sat Jan 13, 2007 7:04 am

Yes, we could tell you. No, we're not going to. (And KAA happen to be pretty good among dvd rippers, which admittedly isn't saying much. Recent KAA anyway.)

laxxy · Post by **laxxy** » Sat Jan 13, 2007 7:14 am

pelican wrote:Yes, we could tell you. No, we're not going to. (And KAA happen to be pretty good among dvd rippers, which admittedly isn't saying much. Recent KAA anyway.)

Speaking about groups, I think there is definitely a problem: you've got the first place occupied by WF/KAA, who made all of 4 episodes of one series together before quitting. There needs to be some adjustment for the # of votes, or # of episodes subbed, or smth like that.