Regarding the recent site issues (502s) and Bayesian averages and ratings histogram

Status
Not open for further replies.
Miku best girl
Admin
Joined
May 29, 2012
Messages
1,441
First of all, happy new year!

We have gained a lot of traffic over the holiday period, and once again became a victim of our own success. The number of users at peak times is pushing the webserver to the limit, and this is the reason for the 502s. An upgrade to the infrastructure is on the way, made possible by your generous support, so keep it up! It should be ready in time for our 1 year anniversary.

Now, onto other new features...

There is now a ratings histogram on the title page, which you can toggle to view. This will give a good idea of how other people have voted.

In addition, Bayesian averages have been added to the title page. This new method of calculating the average is designed to negate the effect of small number of extreme voting on either end of the spectrum. If you're interested, the formula is:

unknown.png


Feedback on the usage of this formula is welcome. The average rating on the site around 7.9.
 
Joined
Jan 3, 2019
Messages
8
Thanks, and happy new year for you too!

I hope there are an option to hide Korean Manhwa / Chinese Manhua and only showing Japanese Manga soon~

Well, basically the same feature similar to those Hentai toggle hide/show.
Like this one suggestion https://mangadex.org/thread/44472
 
Joined
Dec 7, 2018
Messages
32
Would you please add an option to remove "demographic" tag from front page as well? There are a few tag I wish to disappear from my front update page.
 
Member
Joined
Feb 20, 2018
Messages
78
Thanks for the histogram. It's interesting to see the ratings breakdown.

I was wondering how the score would be affected by using the genre or demographic average, but I realized that the scores would probably get inflated by the overall enthusiasm for a genre or demographic...
 
Dex-chan lover
Joined
Jun 22, 2018
Messages
2,109
I was wondering if the 502 Errors were network connection issues (thank you ever so much, Windows 10) on my end or not. Glad to hear that it should be fixed with the new server upgrades coming.

Bit off-topic, but is there any chance of manga search results or listings in general having a little icon (along with views and ratings, ect.) to tell you how many chapters/releases there are? Because there have been multiple times where I've looked at a manga only to find out there are no releases for it yet.
 
Contributor
Joined
Jan 28, 2018
Messages
1,414
Thanks for the more useful Bayesian averages. Looking at the histogram, it seems a like around a quarter of readers rate everything as a 10, limiting the usefulness of the regular average.

Will we be able to sort manga search results using the new scores?
 
Member
Joined
Aug 8, 2018
Messages
10
That's real cool. It would be interesting to see a graph of how the rating of a series change over time. It would show drops in quality or controversial moments.
 
Member
Joined
Dec 28, 2018
Messages
62
I'm one of the new people over the holidays crashing your servers! I gave up on Kissmanga after all the trojans and blocking of adblockers. I understand why they felt they had to go that route. Hopefully you guys wont and I hope to be able to contribute so you can keep it that way. Thanks!
 
Contributor
Joined
Jan 18, 2018
Messages
1,875
Oh thanks for the Bayesian averages and the histogram, it's pretty interesting to look at these numbers mmm...
 
Dex-chan lover
Joined
Jan 27, 2018
Messages
2,430
Thanks for the update!

Quick question, regarding content tags for the H portion of the site, would it be possible to implement something similar to E-hentai/ge-hentai?

As in seperate tags for males and females, and certain tags auto including others, ie. Male:shibari auto adds male:bondage, or female:elf adds Fantasy automatically.
 
VIP
Joined
Feb 11, 2018
Messages
375
The average rating on the site around 7.9.
Neat.

For comparison, the current overall mean for ANN appears to be 7.7637.

I wonder what the site-wide averages are for the other tracker sites like MAL, MU, AniList, Kitsu, Anime-Planet, etc. The value doesn't seem to be as easy to find for these other sites.
 
Dex-chan lover
Joined
Feb 21, 2018
Messages
1,416
The last time i heard bayesian was in college damn i feel so old now its 2019 -_-
 
Member
Joined
Jun 27, 2018
Messages
265
does this calculation consider that the maximum difference in rating is only 9 stars,since you can't rate 0 stars?

To get a decent result the formula should reflect this difference, otherwise higher votes will always outvote the lower ones due to the +1 default rating if you want to vote something as horrible, while rating something as 0 stars (i.e. not rating at all) is not included in the statistics.

By the way, about 80+% of the ratings are rigged either way because of the horrible quality of the manga in general.
If someone votes a bad manga a 10, that's because his subjective enjoyment of the manga overwrote his human sanity and reason (if he ever had that in the first place).

Sure, 99,999% of the readers are not literary critics, but by introducing an algorithm that negates the "voices of sanity" (i.e. the few that actually rate a bad manga badly while others just rate everything they read 7-10) you're actively supporting even more rigging on the rating end.

Might as well introduce an option that removes the rating from the page so we don't have all these moe manga (which are nothing more than a display of modern art created to grob for the money of some derailed otaku) and harem borderline hentai stories listed as top.
 
Aggregator gang
Joined
Mar 17, 2018
Messages
849
I'd like to point out that leaving categories with 0 votes is NOT a histogram. Currently this is some frequency graph which needs to be read with more attention than normal histogram.

Second problem. Is global mean for the site calculated between manga or between all ratings?
 
Miku best girl
Admin
Joined
May 29, 2012
Messages
1,441
The global mean is for all ratings, not for all titles.
 
Joined
May 3, 2018
Messages
86
nice, i though my connection was broken, but this is 502, well thanks for everything
also i'm sure i already finish study about bayesian thing but why i found you again
 
Status
Not open for further replies.

Users who are viewing this thread

Top