|
I got this idea after reading tons of random reviews in IMDB.
I begin with this:
1. Inception is a 8.9 / 10 movie
2. I want to know what those people who rated 1/10 says
So i went on and read them, those people who rate 1/10 seems to rant rather than review movies. So i went on and checked out their other reviews of other movies... which shock me.
most of them rates 1/10 for,
LOTR, the dark knight, batman begins, inception, inglorious bastards, and transformer.
most of them then rate random movies 10/10 when the average is 5/10 or 4/10 simply to pull the rating up
but what shocked me most was, one of the comment he mentioned was this "I feel obligated to rate is 1/10 because i need to bring down the overall rating...."
so he assumes for example, 10 people rating 10/10 + 1 person 1/10 equals to 8/10 which he feels is the correct rating....
translating to how ANN rate their anime, this system is identical to how its done there... there are people out there who just want to mess the actual numbers.
--------------------------------------------------------------------------------------------
So i got this idea:
Take total score -> total reviews : rough bayesian estimate
Accept only -/+ 2.0 from the base average and only use them
Adjusted total score -> adjusted total review : adjusted bayesian estimate
and we shall only use the adjusted bayesian estimate
--------------------------------------------------------------------------------------------
I went further to actually checked random users in Myanimelist and their average anime score compared to actual result. People who usually thinks before writing review and score, those people with 200 over series watched usually has their score withing -/+ 1.0 of the actual average.
This means 2.0 will cover 95% of all the population of people who actually rates properly, and it will ensure more accurate scoring. By all means, kept displaying how many 1/10 those anime has, but as the actual score, ignore them =)
|