dwchang wrote:trythil wrote:This is impossible to achieve with the current opinions system.
Make a suggestion like I did
.
Trythil got to this before me, but he's right: by consistent I meant from person to person. And this is un-fixable since the reviewers are what statisticians call "self-selecting". Meaning they choose which videos they want to look at and review as opposed to being randomly assigned.
dwchang wrote:It's a matter of which you get more of which in turn biases the data in that direction. I feel the "too nice" is that bias and it's shown in our global averages.
There is that bias, but there is another bias that is more revealing. On the first page of the star scale list you'll see 20-40 AMVs that have more than 500 ratings On the last page - or the 2nd to last if the last has less than 100 entries - you'll see exactly...zero...that have 500+ ratings, although one or two are close.
In plain English, the most desired (because people
choose what they want to see), most downloaded, most rated AMVs are the ones with the highest scores. In plainer English, it's a popularity poll. Mainly it's popularity of specific videos, although popularity of creators may factor in to some degree. And a lot of other things factor in as well, so I'm not going to claim that it's purely a popularity contest, but it is mostly that. The numbers are the proof.
This is - allow me to speculate
- why you're thinking that more opinions per video is the solution. The AMVs with the highest number of ratings are actually very good. So, it appears that the rating system is working. But it's only an illusion since the other AMVs - in the middle or at the bottom - are not getting the same treatment.
This wide difference between the rating counts at the top and the bottom is clear evidence that the system has no good statistical basis. And the general rule that more samples are always better is only true in carefully constructed situations.
Again, this doesn't mean the system needs changing. But it should not be assumed to be what it isn't.
