Perhaps it isn’t a secret, I’ve a thing for wondering why individuals can sincerely trust finite point-scale ratings (10 balloons, 3 hearts, 5 clouds).
First, in response to lelangir:
We need standardization, systems of agreement. Concrete ways to convey opinion through the same system [of ratings]. Simultaneously, this leads to disagreement, but that’s a good thing.
Yes, what about this? It’s not that confusing, though lelangir didn’t quite catch my comment. So here is the gist in picture form.
I hate pimp my own creation, but I’ve developed an alternate form of ratings, which attempts to be relavent for users such that their ratings may relate to others, without questioining what so-and-so considers a 10.
Or perhaps rating may be too small to conceal ideals? Or perhaps we need moar 9’s and 10’s to be irrelevent in the world? Like over 4000 votes 10.0! That would be great, thank god imdb does some secret weighing in their calculations.
I truly wonder, is my proposed system/concept too difficult? (Ordering things top to bottom) Perhaps I should bring some kindergardeners both ratings systems and see which one yields better results…
Why should we force titles into another person’s containers (1-10), when there is a viable option to each have our own rating system, which is able to correlate with others. Sure there is some complex summations and equations in the background, but all users have to worry about is:
A>B
A<B
A=B
ah ok I get it now, thanks!
so…say, that in the melative database RRS, I search for “TTGL”. When the results come up, I may then search within that query for “Marimite”. I’d then get 5 hits saying TTGL > MM, 12 hits saying TTGL < MM, or 2 hits saying TTGL = MM.
??
Melative at large can then average these statistics to say “this is the anime that has received X total [greater than] votes.” “this is the anime that has received X total [less than] votes” etc.
that’s interesting, pretty cool too.
lelangir, well that is the gist, but it’s a bit more complicated. Also, I’ve not implemented any specific calculations on the numbers… but what you say is correct.
Well, it won’t actually show you the variations on equivalence, because that is not the sole thing that matters. If you consider say 1000 lists with 1000s items each, you want to know the extent at which they vary in position on the list. Not only greater-than, but also by how much (how many positions), as well.
More importantly, you could take a users’ entire list of ratings and conform them to your own (this is possible by crunching some of the numbers in the backend). This enables no single user’s list to be the right one, but collectively we have something to look at that isn’t absolute.
There are also other complicated matters (not in the front-end) that determine how exactly a user’s list changes the overall system, generally the more dynamic of experience, the greater effect an individual’s ratings can have on the system … this isn’t to say someone can have a list of 1000, create 900 levels and put 100 bogus titles on the first level to sway the system … this sort of thing is accounted for, as are others.
I wandered across the continuum case along a totally different track, a long time ago. I didn’t know that there was an implementation up and running several months before. Shows that I’m perpetually behind the curve.
No no, intro, you were definitely in the right place. I think this list-order rating really struck me when I began to do such things with my anidb ratings down to the 100ths, this was somewhere around Jan 2007. Something told me it was difficult to put two things in the same place, and then to change it later on D: was madness.
I was one of your very first members. Someday, I’ll know how to use your platform, and I admire the relative grading system of melative. :3
Good luck with its end, Ryan!
Michael, hehe, yes … also, you can use the grading atm, I think its in the dashboard.
The main problem with people rating things is that they tend to associate the description of ‘average’ with things that are actually ‘bad’. In any rating system you should expect the average, run-of-the-mill anime to fall at around the half way point (so 5 on MAL). However, because 5 is termed average, people are hesitant to attach that to a series they enjoyed, instead chosing 7 to represent those kind of series. It’s kind of annoying. more annoying is the fact that people’s ratings, when they aren’t divided up into catergories, incorporate both their objective and subjective viewpoints. Ratings are generally done by first looking at your enjoyment of an anime or an episode, and then looking at its merits relative to other anime in terms of things like production and writing. Everyone seems to have a different ratio between these two perspectives, which means a rating is only useful if you know enough about the individual to understand what that means in terms of your tastes.
Another thing that annoys me about ratings is the strange obsession people tend to have with 10/10 or 5/5 being a “perfect” score. Given that the rating should be linear, 10 should appear as often as 1 does in any sensical distribution. That doesn’t seem to happen very often.
Strangely I take ratings quite seriously, and because I have a page that describes what each one means, I’ve tried to stick to the description as exactly as possible.
@washi, you have structure and solidarity. Of course it would be nice if everyone had some structure to their ratings, because they just become less and less useful as more 10 year olds start dishing them out.
About the 10 and 1 appearance, this may or may not be, but what is interesting is what a user’s curve of ratings would look like (Gaussian?). I’m afraid it would not be Gaussian, but as to dealing with such ratings and “standardizing” them across all users … there are ways, but its much more complicated than the proposed system above.
Oh and a reason why 10 and 1 will not appear as regular is quite possibly the issue of selection; it’s not random, users choose what to experience.
>>Oh and a reason why 10 and 1 will not appear as regular is quite possibly the issue of selection; it’s not random, users choose what to experience.
Yeah I suppose that’s true, and that’s the same reason that you wouldn’t expect it to be Gaussian distribution.