39 replies to this topic

[DarkD]

In case my topic wasn't clear, I'm referring to the ratings that people can give an individual manga that has a maximum of 5/5.  

 

Now I am not 100% sure that I'm right in this, I'm hoping I am because there is absolutely no point in even using the score of a manga to judge its quality on batoto.  No manga gets less than 4.  It has to be abysmal to get less than 4.  Almost every comic gets pretty much the same score, so why use it.  If I wanna know if a manga is quality or not, I go to myanimelist and look it up there because they have a good scoring system which accurately represents the quality of the manga.  

 

Take Beelzebub.  Its a great manga, but its not as good as your site makes it look.  High 80s at best.  I would give it an 82/100, but your site gives it a 4.5/5 which translates to a 90/100.  The side story is kinda crappy by comparison, and I would give it a 72/100.  You give the side story a 4.9/5 which translates to a 98/100 meaning your site would put this half assed side story on the same level as legendary manga which is must read such as Berserk, Hellsing, Monster, Death Note, etc.  MAL gives them 83/100 and 79/100 respectively.  

 

Take Maga-Tsuki as another example.  Its a trashy ecchi which isn't quality, it's more a part of a super popular genre where people don't care if it's got a terrible plot.  I would give it a 74/100, MAL gives it a 78/100 you gave it a 4.5 which translates to 90/100

 

The reason for the discrepancy is that with a 5 point scale people have to choose between 4/5 which translates to an 80 which is often too low for what the reader wants so he just gives it a 5/5 which translates to a 100/100.  People are afraid to vote 3/5 because that rating generally means its utterly mediocre and probably won't even interest its own target audience; thus, people give it a 4/5.  In the end, everyone votes either 4/5 or 5/5 and every score on batoto ends up being completely meaningless as a result.  

 

I don't know about other people, but I consider an 80 and above with a genre that's in my general interest range as readable.  70-80 I am willing to do if it's a niche I really like.  90-100 I will give it a shot even if it's considerably out of my interest zone. 

[Kannade]

yeah we need to make the scoring system out of a thousand. if a manga is really bad, we should have the option to rate it ☠/1000 .

Edited by Kannade, 20 August 2014 - 04:12 AM.

[Bobby W.]

The problem is, even if we utilize the score to be 0 to 1000 (for example), most fanboys will rate 1000 anyway. The same reason as to how most manga's got 5 stars. And then haters will vote 1, just like how they'll vote 1 star. There aren't many readers who judges manga's equally and objectively.
In my case, I don't trust the rating system at all. I pick manga through popularity for now (views count).

 

But well, that's just my hypothesis, perhaps it may actually work somehow.

Edited by Bobby W., 20 August 2014 - 05:38 AM.

[svines85]

You mean people actually look at those ratings to decide if they're going to read something? Of all the things I look at to decide if I'm going to give one a whirl, I can't say I've ever considered any type of rating system (on any site) even once.

 

But to each their own, though I've got to say using those ratings as some kind of litmus seems like it'd be really restrictive. People have different tastes, discounting something you'd otherwise try over some cooked up number seems pretty short-sighted.

[Chilled SDK]

An opinion of someone who occasionally uses the rating system now; I really dont see anything wrong with it. If you're around Batoto enough, you most likely will be able to tell the general area for highly rated comics without considering the actual percentage for ratings. A quick look down will be able to tell you whether the manga is there for being ecchi, or any other genre which you may not like. A more detailed system just seems like it would become too bothersome to contribute to for most readers.

[Shrimpeh]

i think a 10 point system would be better imho.

Sometimes i don't vote because i think '3 or 4 is too low, while 4 or 5 is too high'

[DarkD]

You mean people actually look at those ratings to decide if they're going to read something? Of all the things I look at to decide if I'm going to give one a whirl, I can't say I've ever considered any type of rating system (on any site) even once.

 

But to each their own, though I've got to say using those ratings as some kind of litmus seems like it'd be really restrictive. People have different tastes, discounting something you'd otherwise try over some cooked up number seems pretty short-sighted.

The one on myanimelist is actually fairly accurate, just look at my examples, my personal ratings were actually fairly close to myanimelist's.  Personally I would go a step further and ban users from rating something if they consistently give series perfect or 1/100 scores when they are clearly no where close.  

 

a 100 or 1000 point score would be ideal yes.  But 5 is just garbage and you shouldn't even have it on the website if its gonna be something that useless.  It's more misleading than anything.  

 

 

 

An opinion of someone who occasionally uses the rating system now; I really dont see anything wrong with it. If you're around Batoto enough, you most likely will be able to tell the general area for highly rated comics without considering the actual percentage for ratings. A quick look down will be able to tell you whether the manga is there for being ecchi, or any other genre which you may not like. A more detailed system just seems like it would become too bothersome to contribute to for most readers.

 

 

While most of what you said is true, the problem is that if you're using the other metrics to judge the series, then what's the point of having the rating system.  Ratings are also supposed to be a way around the other problem known as ecchi....  It's ridiculously popular and almost all the series which have it as a genre have almost no storyline at all.  

 

Take To-Love-Ru anime, I would give the first anime a rating of 7.8, and the second anime a rating of 6, however how many views do you think that series has?  It's a lot.  Yes a skilled veteran of anime and manga can quickly sort through things like that, but why should everyone be a skilled veteran.  This is the biggest problem batoto has right now and it's so easy to fix.  

 

1) Reset scores to 0

2) Change to 100 or 1000 point scale

3) Done!!!

Edited by DarkD, 20 August 2014 - 08:59 PM.

[svines85]

The one on myanimelist is actually fairly accurate, just look at my examples, my personal ratings were actually fairly close to myanimelist's.  Personally I would go a step further and ban users from rating something if they consistently give series perfect or 1/100 scores when they are clearly no where close.  

 

a 100 or 1000 point score would be ideal yes.  But 5 is just garbage and you shouldn't even have it on the website if its gonna be something that useless.  It's more misleading than anything.  

Spoiler

 

 

 

 

While most of what you said is true, the problem is that if you're using the other metrics to judge the series, then what's the point of having the rating system.  Ratings are also supposed to be a way around the other problem known as ecchi....  It's ridiculously popular and almost all the series which have it as a genre have almost no storyline at all.  

 

Take To-Love-Ru anime, I would give the first anime a rating of 7.8, and the second anime a rating of 6, however how many views do you think that series has?  It's a lot.  Yes a skilled veteran of anime and manga can quickly sort through things like that, but why should everyone be a skilled veteran.  This is the biggest problem batoto has right now and it's so easy to fix.  

 

1) Reset scores to 0

2) Change to 100 or 1000 point scale

3) Done!!!

 

"Accurate"........how? I mean, what are you basing this on? You've got some systematic formula? Some assignment of values? Does everybody else know about it and use it too? Does everybody else agree with your personal assessment on each and every title? This is all very, very subjective stuff, man.

 

Look, no offence meant, I'm not specifically downing your idea, I'm just voicing an opinion......... but like I said I consider "ratings" as some subjective, cooked-up number that I really don't assign hardly any relevance to. 1 out of 5 or 1 out of 1,000,000, it still wouldn't get more than a glance from me  

 

 

Personally I would go a step further and ban users from rating something if they consistently give series perfect or 1/100 scores when they are clearly no where close.  

 

Haha! Dude, come on, I think you're taking this way too seriously.

[Gendalph]

I've thought about this long before you did.

tl;dr: ratings are useless - people are biased, they never give low rating.

Average rating is 3.73*
Median rating (for my data, end of May, I thing) is 4.03*
Lowest you can give is 1*, highest - 5*, so fair median should be close to 3*.

There are:
about 400 comics that have no rating;
~900 comics that have rating less that 3*;
~3000 comics that have rating 3.x*;
~6.5-7k comics that have rating 4.x*;
less than 50 that have 5* rating (most of them have 1-3 votes).

Yes, this sucks, but this is how ALL ratings are. There is no way to fix this except for rank based on reviews made by critics. Professional critics. And you'd need at the very least 3 reviews to get somewhat adequate ranking.

- What about views?
- They have some value, but those are just views, they mean nothing. At all. Every page view (even bot) will add +1 to views.

- Hey, wait, we have follows!
- And? Let's look at Psyren, Pluto and Beelzebub (just cuz I like those):
http://myanimelist.net/manga/3438/Psyren
https://www.mangaupdates.com/series.html?id=19081
http://myanimelist.net/manga/745/Pluto
https://www.mangaupdates.com/series.html?id=226
http://myanimelist.net/manga/10010/Beelzebub
https://www.mangaupdates.com/series.html?id=41265

            rating   views    follows   votes   views/follows   fol/vot   vot/viw*1000   fol/viw*1000
Psyren      4.5      385.3k   1620      436      237.84          3.72     0.88           3.27
Pluto       4.59     113.4k   362       88       313.26          4.11     0.77           3.16
Beelzebub   4.5      39.2M    10141     1006    3865.50         10.08     0.0257         2.59

Does having 100-300 times more views make Beelzebub more amazing than Psyren or Pluto? Hardly.
Does Pluto suck just because it has less follows? But hey, it's OLD. And psychological. Those don't have as much follows. Psyren is maybe not as good as Beelzebub is, but hey it's just older (check MU/MAL - they have quite close rankings).

- Then what about votes?
It shows how many users ranked this series. And it's more accurate... Maybe... For current series... But not so much for jewels like Pluto! Which is ranked higher on both MU and MAL.

- Ok, let's use two parameters!
- Sure. Which?
Views to follows: Yeah, newer series have a huge advantage there - not fair.
Follows to votes: mhm, again...
Votes to views: yea, but is Lost Man by Kuji Shinnosuke that good? (it has 8.5) First old title I saw in this rating is Ares (#162 with 2.12), ToG is #9467 with rating 0.06 (my data is from before takedown).
Follows to views: look at numbers. Are you content with them? I'm not.

Bayesian average? Pft, it works only in fairy tales:
- if you set base rank too high, nobody's gonna vote down trash;
- if you set it too low - newly uploaded series will have low rating;
- if you set minimum votes number too low - yea, it will work, but was there any necessity to use it?;
- too high - everything will have rating close to average.

There is a more pessimistic method (I've done some research and tests). Does it solve the issue? No: it mixes everything but popular series with dirt.

I know no proper way to rank neither older series, nor brand new. I've thought about it, long and hard, but haven't found any viable solution.

[Chilled SDK]

While most of what you said is true, the problem is that if you're using the other metrics to judge the series, then what's the point of having the rating system.  Ratings are also supposed to be a way around the other problem known as ecchi....  It's ridiculously popular and almost all the series which have it as a genre have almost no storyline at all.  

 

Take To-Love-Ru anime, I would give the first anime a rating of 7.8, and the second anime a rating of 6, however how many views do you think that series has?  It's a lot.  Yes a skilled veteran of anime and manga can quickly sort through things like that, but why should everyone be a skilled veteran.  This is the biggest problem batoto has right now and it's so easy to fix.  

 

1) Reset scores to 0

2) Change to 100 or 1000 point scale

3) Done!!!

Arent you generalising the perspectives of others a little too much? It seems more so that you are trying to suggest to change the rating system more to something which will never exist. Ratings are based on people's perspectives on the series. If a person particularly likes a genre and the manga is suited to the needs of that person, he will most likely rank it high. Not all ranked high manga will be to your own tastes, that's why other metrics such as genre are needed to identify if the manga is something you'd like to read.

 

Also, have to say a 1000 point scale is absolutely ridiculous. At max a 10 point scale should be used.

[bastek66]

Maybe do like youtube and change it to Like and Dislike.

[DarkD]

 

"Accurate"........how? I mean, what are you basing this on? You've got some systematic formula? Some assignment of values? Does everybody else know about it and use it too? Does everybody else agree with your personal assessment on each and every title? This is all very, very subjective stuff, man.

 

Look, no offence meant, I'm not specifically downing your idea, I'm just voicing an opinion......... but like I said I consider "ratings" as some subjective, cooked-up number that I really don't assign hardly any relevance to. 1 out of 5 or 1 out of 1,000,000, it still wouldn't get more than a glance from me   

 

 

 

Haha! Dude, come on, I think you're taking this way too seriously.

Accurate in that if something gets a 7-8 on MAL I have a consistent idea of what to expect quality wise.  Take Mahouka anime, it got in the high 7.  After reading that I knew exactly what to expect and I got exactly that.  My expectations were that it was a third rate anime which you will like if you can identify with one of its many niches.  I could identify with a couple of them so I was able to watch about 10 episodes.  Exactly what I expect from an anime with a rating of 7-8.  I would chalk this up to opinion if it was just one off time, but it's not.  I am consistently able to rely on myanimelist's ratings for what to expect from a series.  That's not subjective anymore if I can get consistent results out of it.  

 

 

Arent you generalising the perspectives of others a little too much? It seems more so that you are trying to suggest to change the rating system more to something which will never exist. Ratings are based on people's perspectives on the series. If a person particularly likes a genre and the manga is suited to the needs of that person, he will most likely rank it high. Not all ranked high manga will be to your own tastes, that's why other metrics such as genre are needed to identify if the manga is something you'd like to read.

 

Also, have to say a 1000 point scale is absolutely ridiculous. At max a 10 point scale should be used.

 

It's more that ratings as they are used right now are useless as you can see by everyone's comments.  No one uses them.  They have zero credibility.  However, MAL has cracked the code, their ratings are useful.  So I am trying to implement that system here.  Here's how it works, everyone has certain series which they like, I like fantasy and supernatural series.  So I search by genre and then I order them by their ratings.  Then I read through the descriptions and pick out the ones near the top that interest me.  It works.  

 

100 point scale is actually fine, but 10 point isn't much better than 5 point.  MAL does 1000 point.  

 

Maybe do like youtube and change it to Like and Dislike.

 

You're missing the point.  Youtube has a half assed search mechanism targeted at the everyday viewers, batoto has a first rate search system targeted at the hardcore computer users.  The reason it matters here is because you can combine the rating with other criteria and get better results from it.  

[Bobby W.]

Here and there, I'll eventually rate my favorite series with highest scores anyway. So yeah, I can't really see how it's going to work here anyway. Well, let's see the other's perspective first...

[svines85]

Spoiler

 

Accurate in that if something gets a 7-8 on MAL I have a consistent idea of what to expect quality wise.  Take Mahouka anime, it got in the high 7.  After reading that I knew exactly what to expect and I got exactly that.  My expectations were that it was a third rate anime which you will like if you can identify with one of its many niches.  I could identify with a couple of them so I was able to watch about 10 episodes.  Exactly what I expect from an anime with a rating of 7-8.  I would chalk this up to opinion if it was just one off time, but it's not.  I am consistently able to rely on myanimelist's ratings for what to expect from a series.  That's not subjective anymore if I can get consistent results out of it.  

 

 

 

It's more that ratings as they are used right now are useless as you can see by everyone's comments.  No one uses them.  They have zero credibility.  However, MAL has cracked the code, their ratings are useful.  So I am trying to implement that system here.  Here's how it works, everyone has certain series which they like, I like fantasy and supernatural series.  So I search by genre and then I order them by their ratings.  Then I read through the descriptions and pick out the ones near the top that interest me.  It works.  

 

100 point scale is actually fine, but 10 point isn't much better than 5 point.  MAL does 1000 point.  

 

 

You're missing the point.  Youtube has a half assed search mechanism targeted at the everyday viewers, batoto has a first rate search system targeted at the hardcore computer users.  The reason it matters here is because you can combine the rating with other criteria and get better results from it.  

 

Hey, if it works for you that's great. I've known/know plenty of people who won't touch a book that doesn't show up on the best-seller list and are perfectly happy limiting themselves like that. My opinion is that it's an exceedingly shortsighted method of evaluating whatever medium one's talking about (if indeed one is basing their choices to any degree on it.)

 

I'd think Gendalph's post of many many numbers probably epitomizes this discussion............you're just playing with numbers here man. You seem to think there's some "perfect", "scientific" method to tell you what to try rather than just evaluating what you can find and then giving it a try. 

[Purple Library Guy]

Lowest you can give is 1*, highest - 5*, so fair median should be close to 3*.

Doesn't follow at all IMO.  For instance, if we assume scanlation groups aren't composed of total idiots, we might conclude that average ratings are high because scanlators avoid putting effort into scanlating crap and so the average quality of scanlated manga is high.  Or the same might be true of publishers in the first place (although that leaves out doujins).

Or we might say people are a bit nice and that's a good thing.  Also there's nothing that dictates the quality of artistic works has to be in an even distribution across the conceptual space of how good/bad it is possible to be, or in a bell curve radiating from the middle.  It's perfectly plausible (even allowing the notion of absolute taste such that quality scores are genuinely assignable in the first place) for the distribution of works to be lopsided upwards relative to the aesthetic midpoint between the very worst points and the very best.  Of course even talking about this points up the absurdity of the whole construct; you can't say what numbers "should" be on a scale that's arbitrary and meaningless to start with.

 

Ultimately, the average score is whatever it is, and any minimal meaning you might extract will have to be based on variation from that average.

Meanwhile, of course the smart way to pick manga is to look at my list of favourites!

[roch]

tbh my way of knowing whether or not this a Manga is worth a read is the number of follows, the tags and the comments (on the lastest chapter & the first bunch of comments when the manga started) .. i never look at the ratings...

[kukin]

why a 5 point score is useless

 

More like

 

point score is useless

[red wolverine]

we could also have a bayesian average, like mangaupdates

https://en.wikipedia.org/wiki/Bayesian_average

 

or use other methods like what percentile the score is in, or give a normalized score between the lowest and highest overall (site) scores

[algormortis]

Maybe do like youtube and change it to Like and Dislike.

I agree with this.

 

It would be quite useful to see a flat number of likes vs dislikes, with the red/green percentage bar showing the difference.

 

I'm pretty sure the way people go about rating manga is as follows (and I'm generalizing quite a bit here):

If they like it, they give it 5 stars, because they feel the current rating isn't high enough. If they don't like it, they give it 1 star, because they feel the current rating is too high.

 

Alternatively, having individual ratings for story and art would also be helpful. People have their own interpretations of what the 5-star rating means. I personally feel that having 4.8+ ratings on ecchi manga is a bunch of crap, because very few of them have a good enough story to merit such a high rating.

Edited by algormortis, 07 September 2014 - 01:51 PM.

[red wolverine]

Maybe do like youtube and change it to Like and Dislike.

I agree with this.

 

It would be quite useful to see a flat number of likes vs dislikes, with the red/green percentage bar showing the difference.

 

If we take 1-2 stars as dislike, 3 as neutral, and 4-5 as like, we could get a percentage bar  and like vs dislike number from the existing data

Edited by red wolverine, 07 September 2014 - 02:37 PM.