Recent Ratings

[WindOcean]

I suggest adding a second rating that only takes into account recent ratings.

Reason:
Often a manga starts strong and the first few chapters are rated very well.
Then after a few more chapters the author drops the ball or the quality worsens.

Why?:
People tend to rate things they have read. They tend to read the manga till the last chapter and then rate it. If more chapters are added the recent rating should reflect the overall manga quality better.

Negative:
Sometimes there aren't enough ratings to be accurate.

Alternative:
Add ratings over time for people to identify trends, like is the manga getting better or worse?

 

[Roler]

🤔 Maybe a chart would be better.

 

[undeed]

as someone who once devised slightly similar idea, here's my take

People tend to rate things they have read.

me yes

They tend to read the manga till the last chapter and then rate it.

me no

If more chapters are added the recent rating should reflect the overall manga quality better.

Wait, overall manga? isn't that then just... the rating?

And the reasoning "tend to read the manga till the last chapter and then rate it" contradicts with the "only takes into account recent ratings"

Negative:
Sometimes there aren't enough ratings to be accurate.

Now how often do you see a manga with a score lower than 5 here?
The calculation isn't perfect yet.

Alternative:
Add ratings over time for people to identify trends, like is the manga getting better or worse?

Basically it boils down to the "alternative".
Which is per chapter scoring.
Tho how hard would it be to implement? I don't know.

 

[undeed]

However, I would still prefer one overall score if the calculation could be improved.

 

[Basic_human]

yeah rating system need to be improved cuz ive been deceived by the rating a lot

8 or almost 9 manga turn out to be mid is frustrating

 

[Blocky_Mudkip]

yeah rating system need to be improved cuz ive been deceived by the rating a lot

8 or almost 9 manga turn out to be mid is frustrating

Maybe you just have bad taste

 

[RogueKitsune]

Ratings are inherently biased as users tend to rate titles they read and only read things they think they will like. This results in titles having a higher rating than they would if every user were to rate it.

 

[undeed]

Ratings are inherently biased as users tend to rate titles they read and only read things they think they will like. This results in titles having a higher rating than they would if every user were to rate it.

Are you talking about the scores in mangadex or manga/media in general?
Because I don't think that was the case here in mangadex.

I've noticed this since a while back.
You can find a title that is still fresh and no one has scored it yet, and it still gets an 8 (diplayed).
Because that's how it's calculated currently.

I once found a title with only like 2 people giving it a score of 1, and the displayed score was still 8 (can't really remember).
So instead of looking at the average score, it's better to look at the details of the score distribution.

or,
I suggest to just display the average score

 

[Tony_Mushah]

8 or almost 9 manga turn out to be mid is frustrating

Yeah... That's just the part of the whole experience of reading manga here on Mangadex.

 

[BraveDude8]

Are you talking about the scores in mangadex or manga/media in general?
Because I don't think that was the case here in mangadex.

I've noticed this since a while back.
You can find a title that is still fresh and no one has scored it yet, and it still gets an 8 (diplayed).
Because that's how it's calculated currently.

Bayesian average. It smooths out ratings at low vote counts by adjusting them towards the site average to avoid massive swings where three people rate an new title 10/10 and it immediately hits #1 on the site. The more ratings a title has, the less affected it is.

The reason why brand new titles with no votes show a rating of 8.02 is because the average rating of every single title on MD is 8.02.

8 or almost 9 manga turn out to be mid is frustrating

According to the MD userbase, 8 is mid. I realise it's not intuitive, but it nicely demonstrates how people basically only use 6-10 as a rating scale and there's not much we can do about it.

 

[undeed]

Bayesian average. It smooths out ratings at low vote counts by adjusting them towards the site average to avoid massive swings where three people rate an new title 10/10 and it immediately hits #1 on the site. The more ratings a title has, the less affected it is.

The reason why brand new titles with no votes show a rating of 8.02 is because the average rating of every single title on MD is 8.02.

I know what Bayesian average is for, I mean, there are people who are clearly said it's 10 or nothing.
I just think that maybe it's better that it's not the one displayed.
Maybe use the bayesian as the sorting method for the site, but either display the normal average or put N/A if it's under a certain threshold.

 

[Basic_human]

According to the MD userbase, 8 is mid. I realise it's not intuitive, but it nicely demonstrates how people basically only use 6-10 as a rating scale and there's not much we can do about it.

Hmm it's hard, were that the case using 5/5 scoring system could work, we wouldn't get the perception of 8 is good score, but we get 4

Since it's mostly more of reader psychology and mental the mods can't really do much huh. Really have it hard huh

 

[BraveDude8]

but either display the normal average or put N/A if it's under a certain threshold.

The entire reason Bayesian averages are implemented is to avoid showing wildly misleading raw ratings in this exact scenario. One person rating a new title a 1 would immediately bury that title in terms of public perception.

 

[WindOcean]

I think, chapter based ratings just are too much of a hassle for me as a reader. Especially when reading hours on end.

A graph with reviews like in Steam, and recent ratings like on Steam, might just do the trick. But that is only a guess.

Regarding ratings: I always felt that ratings on sites like proxer had more meaning, as a score there always has a relation to a long and thought out review. The reviews also can be rated by other users, meaning that the top review is most likely the most accurate one by Democratic selection.