Anime Statistics

Want to help out? Need help accessing the AniDB API? This is the place to ask questions.

Moderator: AniDB

Locked
ncmaothvez
Posts: 8
Joined: Mon Sep 12, 2005 5:37 pm

Anime Statistics

Post by ncmaothvez »

Continued from this post.

I'd like to see statistics for anime ratings and number of votes but I don't think it's that much fun looking at graphs that simply show the actual ratings and number of votes. Instead I find it more interesting if the graph shows a weighted rating based on popularity. I mean, which anime is more popular; the one rated 9.5 with 10 votes or the one rated 8.0 with 750 votes? The way I see it the one rated 8.0 is more popular because the one rated 9.5 is most likely somewhat over rated (due to its low vote count). The more votes an anime has the more accurate the rating becomes.


As an example, here's how I've been calculating my own stats so far:

Each day get a list of all animes for which both of these two conditions are true (this is just to limit the number of animes in the list):
  • Rating >= 7.5
    Number of votes >= 750
From this list, find the anime with the highest number of votes. This number is Vmax.

Then, for each anime in the list, calculate a wheighted rating like this:

Code: Select all

Rw = R * (V / Vmax)

Rw is the wheigthed rating
R is the anime's rating
V is the number of votes for the anime
Vmax is max votes as described above
Then plot these wheighted ratings in a graph.

The nice thing about this method is that the same graph shows two things:

1. The difference between Rw for any two animes on a given date shows how much more popular the one anime is relative to the other.
2. The Rw delta between two (closely located) dates for a given anime shows how 'hot' that particular anime is at the moment; the 'hotter' it is the higher the delta.

This stats calculation method may be somewhat flawed depending on how one sees it: this method assumes that an anime is more popular than another one if it has more votes than the other. This may not allways be true though; it's possible that the longer an anime has been listed in the database, the more likely it is that the anime has a higher number of votes than an anime that has been in the databse for a shorter time. So statement #1 above may or may not be true depending on how one sees it. For the top 10 to 20 animes I think this method is accurate though.

Perhaps there's a better way of calculating the wheighted rating. Maybe the number of days an anime has been listed in the database could be included in the formula. The release/airing date shouldn't be used though since that date may be much older than the date the anime was added to the database.

/N
exp
Site Admin
Posts: 2438
Joined: Tue Oct 01, 2002 9:42 pm
Location: Nowhere

Post by exp »

well,

I guess I'll leave the thinking about the right way to use the values to calculate a score to our math/statistic freaks out there.
i failed miserably in my last attempt, see:
http://anidb.info/perl-bin/animedb.pl?s ... o=hotanime

The first thing we should work out is which values may be usefull in the future and should therefore be collected now.

The current list is definitely not complete, it doesn't even contain vote count or rating.

current values (per anime, per day):

Code: Select all

- number of episodes
- number of files
- number of groups in groupinfo box
- number of users with at least one file in mylist
I guess other candidates would be:
- perm/tmp vote count?
- perm/tmp rating?
- review count?
- review rating?

what else?

BYe!
EXP
Locked