Totally (un)Biased Ratings on Mangadex

[OisE]

In my opinion, there are a lot of popular titles that don't deserve their good ratings. So I am curious :

Is it easy for a title to gain a great rating score on Mangadex ?
Or, am I just too harsh on my ratings/judgement ? (might be both)​

 

[bigtiddyoneesan]

usually 7 or 8 for works i thought were well-made and enjoyed. i think the rating system works in general for finding good manga, but it can also be somewhat circumvented by non-controversial, feel-good manga that people rate 10's in masses.

 

[PantsMan]

I've gotten increasingly lazier over time when it comes to scoring manga on my own list: anything I particularly like and would easily recommend gets a +, series I love get a ++, and anything I particularly dislike to a noteworthy extent gets a -. Anything else I don't feel very strongly about gets nothing.

Ultimately ratings like these are just there to make it easier for me to sift through these lists when looking back over the years at what's stood out to me. It also lets me sidestep that stupid "Man, this series is my favourite, but it's not absolutely perfect so I'll give it only a 9/10" bullshit. Honestly, the level of granularity of choosing whether an average series sits between 5-7 or a bad series between 1-4 does nothing to me.

 

[tristan9]

Is it easy for a title to gain a great rating score on Mangadex ?

The short answer is yes.

The long answer is yes, and you can actually find that the average site-wide rating is around 7.9.
When a new title is added to MangaDex, its initial rating is based on the average rating on the platform (to avoid a single 0-rating causing the title to be rated 0, or conversely for a single 10-rating making it be rated 10).
This is essentially the idea of us defaulting to the bayesian average for these:

Now you might expect the average rating to be a 5/10 in general, but that's not quite correct, for a few reasons:
1. People filter what they even read to begin with, so they usually only read what might interest them, rather than just about anything
2. People tend to bother rating titles only if they really like them or really hate them (most titles have a lot of 8+ and 3- ratings, rarely much in-between)
3. People tend to rather just unfollow and move on from something they just got bored of, rather than bother rating it as some low-average (think the 2-5 range); it's just not worth the effort

This all-in-all contributes to a very skewed rating, which is pretty much always present in any user-driven rating system.

As a platform we then have two choices:
1. Normalize around the website's average rating (ie what we do atm)
2. Normalize around 5/10

The problem with the #2 here is that when the average is 7.9, the "impact" of a 10/10 compared to that of a 0/10 is much lower (mathematically speaking). So if we did #2, you'd hardly see anything rated above 7/10, even if it has an overwhelming amount of 10/10 ratings, for example.

 

[aknighttoremember]

Anything below 6 I would consider fairly enjoyable but not worth mentioning.

If I enjoy it - 7/8
If I think it's award winning worthy - 8/9
If I'm desperate enough to wait it weekly + all the above - 10

Not really harsh with my ratings.

 

[Angry_Panda]

This problem seems pretty universal to online rankings. Simply put, the biggest problem: if I don't really like or dislike something, what's the motivation for me to take the fraction of a second to give it a ranking?

Looking a bit deeper at the matter, as an illustrative aside (which I'll loop back to MD), I have a collection of several thousand LPs which I have cataloged on Discogs, and for the classical titles (which are about 2/3rds of my collection) I've entered my rankings on the 1-5 star scale they use. Like MD, a 0 is unranked - the lowest possible ranking is a 1. These rankings are public, so my rankings get averaged with everyone else's to determine the community rankings for each title, though I can see my individual rankings in my personal collection.

The problem with relying on the community averages comes in how someone decides what criteria to use for each ranking. I personally am using those for the quality of the release from an audiophile standpoint, so I can lay hands on the best recordings of any given work. A 5 is an exceptional performance, a high-quality recording, and an excellent pressing; 4 would be above average in at least two of those; etc. I only have one or two discs that I've ranked a 1 - why would I own bad records with bad music? - and as I've gotten more experienced, I've started searching out certain groups, labels, etc. to improve my chances of getting 4 or 5 star releases when I go to the local record store. This means my average ranking is up to about a 4.25 at this point. Am I ranking too high, or is there simply a selection effect happening?

Many people seem to give rankings based on whether they like the album or not. So a disc I might think is a 5 could have an average of 3 because most people don't care for the music, or a disc with a 4.5 average might be a 3 in my list because the recording quality and pressing are sub-par, even though the music itself is excellent. This gets amplified a lot for something like classic rock - which I don't rank - to the point where just about every significant album of the '70's has above a 4.5 star average. (I'm sorry, but Atom Heart Mother and Presence just aren't that great.)

Coming back to MD, am I (and a thousand other people) ranking a work based on my response to it - this manga has "wholesome fluff" or "compelling characters" or "BOOBA!" so take my ratings points! - or on the art, or on the translation, or...? So the problem becomes a matter of a single dimension being used a number of different ways by people, compounded by a reluctance to rate something 'in the middle' (note the tendency to skew toward the outsides of the range rather than the middle on many titles) - hyperbole made manifest, in a way.

TL;DR - if you're actually thinking about the ratings, you're not doing it wrong. But don't expect everyone to follow suit.

 

[EOTFOFYL]

Humans are biased,
Some things don't get enough love,
Others get too much.

 

[sterven]

Yup it's real easy to gain rating not much you can do about it.

 

[pandascepter]

all i can say is 10

and my shit are low effort, so it got 6 wheeeze