Should the sites like Scribble Hub use the 5-star rating system?

[Valmond]

Okay-okay, Val got the perfect solution. Everyone else should get 1 stars, that way they are all even. And Val should get 5 stars, that way he is always above. Everyone wins.

 

[Anonjohn20]

Okay-okay, Val got the perfect solution. Everyone else should get 1 stars, that way they are all even. And Val should get 5 stars, that way he is always above. Everyone wins.

There's no Smut in your story. -5 stars in SmutHub.

 

[Valmond]

There's no Smut in your story. -5 stars in SmutHub.

Hey, it is still a 5.

 

[Prince_Azmiran_Myrian]

I'm going to be serious for a second here and reply to this... reply... take. The reason why Steam handles the situation well is because no one looks at likes and dislikes. The first thing you see is a freaking video, a trailer of the game. Sometimes it's a trailer, sometimes it's straight up gameplay video. There are also screenshots. Comparing Steam to SH doesn't work. You don't have a video that gives a reader almost a full understanding of wether they'll like the novel or not. It doesn't work like that. And synopsis doesn't help either. No one looks at synopses in Steam. People watch videos and look through screenshots.

This is why we would use ai to generate memes based on your story. Then readers can judge based on the quality of the memes.

 

[beast_regards]

This is not true; I've reviewed products that I bought elsewhere, so Amazon had no way of knowing if I owned the product.

I thought you could review the products you brought on Amazon, not through other vendors. They could and they do keep track of your purchases.

I assume you meant rating, not reading, but it's also not true; it needs to be on your reading list.

This is largely irrelevant, you could add any story on the reading list.

That would lead to more vindictive authors. They are already whining that they want negative reviews removed; imagine if they started whining that negative reviewers should be banned.

Not necessarily.

The negative reviewers they hate would have less power to influence the final verdict as they have only one vote, not counting the alt-accounts.

In the 5-star-rating system, however, the writers are under much larger pressure to reach certain value (and once again, math is on the site of the opposition) while under like-dislike system, there is no such thing.

Instead of going on the lowest possible rating, they would just hit dislike.

(1+5+5)/3=3.6
vs.
Negative+positive+positive=mostly positive

Actually, it may lessen the pressure instead.

Then you know there are five people who liked your story, and five disliked: You will clearly see the views are somewhat balanced and the five haters didn't take that much from you. After all, it should look achievable to get one more like.

However, if you see your story is rated 3.0 and your story needs to be 4.5 you would feel it a more of the struggle to get it there, especially after the other values join in, like twos and fours.

 

[unlaumy]

On a more serious note, I think you guys are forgetting another type of raters.

Trolls and honest raters, fuck them. How about we talk about people who have different priority for the average rating and their personal rating?

Let's imagine a series with 4.8/5 average rating. Someone feels it deserves 4/5. In your ideal world (in SH's current rating system), it should be easy, they should just rate it 4/5, according to their personal rating. What if they think average ratings is more important than their own experience? What should they do then, to make the series fall into 4/5 faster?

That's right, by picking 1/5. In an ideal world, 1/5 represents the absolute shit and the shittiest. But this reader doesn't feel that way, for them average rating is the best representation, and they just dislike the series for being overrated. And 1/5 rating is the fastest way to decline it.

Are they a troll? (Maybe) A weirdo? (yeah) An asshole? (Maybe)

You should also know that Steam makes a 'summary' of people's collective likes and dislikes. Overwhelmingly Positive and so on and so on. In your ideal world, readers don't have to bother with worse rating system(s) by just choosing only like/dislike, but what if, like I said, these readers only care about 'average rating'? They see a series with a too much good ratio, so they dislike dump it until it finally reaches the Very Negative like they want.

So any types of [personal] rating scale becomes less significant when the average rating is publicly shown. That's just the way it is.

So what can we do? Either we fully scrap rating system alltogether, or actually adopt [bell curve] grading system for SH (yes, I was being serious previously).

 

[RepresentingWrath]

Ok, hopefully I will make one last reply here, and I would act calm and collected.

You know why I dislike these kind of conversations? Because people live in their wet dreams rather than reality. The fact of the matter is, this conversation will lead us to nowhere. It's far from being first, second, or even thirsd thread about it. If you were to make a feature request asking to change rating system it won't be first, second, or even third thread about it.

Back to the topic. People keep living in their wet dreams, thinking how this system is better, that system is better, they know how to do better, etc. But the fact of the matter is, we already have a system, it won't be changed, use it. But they outright refuse to use it. They just refuse to utilize it. It doesn't matter what you think is better, you only have this system, please use it so it can work. But they don't! They sabotage it by not using it as they continue to living their wet dream of a different system.

This is the problem. And I understand to them it's not a problem at all. They want a system that is simpler, that is harder to break by not utilizing it properly. You know how every tool is a hammer? Yeah, they will keep using every tool as a hammer untill they don't get the hammer. They don't care other people might want different tools, all they want is hammer.

And they will act as if a 5-star rating system is rocket science. Extremely childish behavior, hiding behind phrases like, "Oh I don't know what those 5 stars means, so I won't use it!" Which is obviously a childish behavior. They saw a 5-star rating system at least once in their lives, and I'm giving them a huuuge discount here by sayin they saw it once. We are surrounded by similar rating systems when we look up game ratings, movie ratings, and so on. We are constantly subjected to them, but for some reason when it's their turn to use it they don't know what it means.

Another reason why it's not a problem to them is because they are extremely selfish. They only care about their feelings and enjoyment. A lot of people here say how, "I only give 5-star rating to stories I like, because I'm not a douche." It's a very good position to have, but it's also extremely narcissistic and harmful to smaller authors who try to make their novels work. Again, readers or authors with established big fanbases here simply don't care about smaller authors.

They don't go out of their way to give small authors 5-star rating. Should they? No, they shouldn't. But the thing is, they inflate ratings of bigger stories. If a small story has one 1-star rating and four 5-star ratings, it will be rated lowerd than a story that has two 1-star ratings and thirty 5-star ratings. The math isn't the point here, I hope you won't get nitpicky here. So they actively ruin any chance of small authors to get higher on certain search results, heck maybe even on trending, perhaps ratings are take into equation. But they don't care. They want their likes or dislikes, and they want to feel good about not being a douche.

I will reiterate this point, we don't live in a wet dream, we live in reality. The reality is, SH doesn't have a like\dislike system. So you harm those authors. Actively harm them. And even with like\dislike, same problem will happen. Not only to small authors, but also to authors who write something unusual. For example authors who use NTR in their stories(HAHA LET'S PICK THIS ONE FRAGMENT TO MAKE A JOKE AND IGNORE EVERYTHING ELSE). If you write NTR you will almost certainly get ratioed.

Yet no one cares about it. People only think about their feellings, people only think about making themselves feel good. They don't care how good or bad it is for the reader. They don't care about other authors. All they want is change their own ratios.

So yeah... Kinda lost my train of thoguhts, but I feel like this is a calm and reasonable reply. At the end of the day, the problem of 5-star rating system isn't 5-star rating system, isn't moderator John, RR spy agents, or whoever or whatever else. The problem of 5-star rating system are childish narcissists who refuse to utilize it, live their wet dream instead of facing the reality, and only think of themselves.

 

[RiceballWasTaken]

On a more serious note, I think you guys are forgetting another type of raters.

Trolls and honest raters, fuck them. How about we talk about people who have different priority for the average rating and their personal rating?

Let's imagine a series with 4.8/5 average rating. Someone feels it deserves 4/5. In your ideal world (in SH's current rating system), it should be easy, they should just rate it 4/5, according to their personal rating. What if they think average ratings is more important than their own experience? What should they do then, to make the series fall into 4/5 faster?

That's right, by picking 1/5. In an ideal world, 1/5 represents the absolute shit and the shittiest. But this reader doesn't feel that way, for them average rating is the best representation, and they just dislike the series for being overrated. And 1/5 rating is the fastest way to decline it.

Are they a troll? (Maybe) A weirdo? (yeah) An asshole? (Maybe)

You should also know that Steam makes a 'summary' of people's collective likes and dislikes. Overwhelmingly Positive and so on and so on. In your ideal world, readers don't have to bother with worse rating system(s) by just choosing only like/dislike, but what if, like I said, these readers only care about 'average rating'? They see a series with a too much good ratio, so they dislike dump it until it finally reaches the Very Negative like they want.

So any types of [personal] rating scale becomes less significant when the average rating is publicly shown. That's just the way it is.

So what can we do? Either we fully scrap rating system alltogether, or actually adopt [bell curve] grading system for SH (yes, I was being serious previously).

A bell curve would be a flawed model because in reality, the points of the bell curve are subjective and we cannot assume ceteris parabus

 

[owotrucked]

The approximate order of operations for *Patent Pending* (JK) "TruescoreSR* (Truescore Story Rating) is as follows for any new score being put up:

1. Adjust Author score based on their nominal score range - Examples: If they have 7 reviews all in the range of 1-3, adjust scores to be 1-5. If all reviews are the same score, they all default to that score, and a heavy weighting penalty is applied (maybe adjust weight down by 50% to start).
   
   
2. Apply Genre tag average weighting to rating - Examples: NTR underperforms the average site rating across all NTR stories, as such, scores are adjusted upwards somewhat. Isekai overperforms, and is lowered slightly. This is averaged with a weighting taken by reverse calculating averages across stories every so often (perhaps once a month).
   
   
3. Edit: Apply an account score weighting based on Genre accuracy as well, some people are better at rating certain genres than others after all (perhaps they keep reading things they hate, idk)
   
   
4. Apply review count weighting accordingly - Set arbitrary review limit to reach full weighting, and adjust accordingly (lets say 10 ratings gets you to full weight on this variable, and you get 10% more weighting for each rating until reaching it)
   
   
5. Apply account wide accuracy rating - Arbitrary variable multiplies your rating accuracy vs the average rating. When you're off by too much on multiple stories it applies a small penalty to your account's ratings in the future. This is also weighted based only on stories that have Green Quality Truescore ratings (which indicates a 90% certainty in score based on types of reviews and total count.)
   
   
6. Periodically adjust and tune parameters provided in these different variable spaces until all numbers work well across genres. With multiple weightings and variables applied on both account and genre levels, it should be easily possible to fine-tune the numbers to encourage good review practices.

Truescore accuracy would be determined based on the accuracy rating of the accounts that are giving reviews (based on their downweight % versus maximum), and averaged with an inverse rule to bring accuracy up with more account reviews.

That's the gist of my idea.

I advocate a completely different approach:

The goal is not to deduce a single truescore, but separate ratings into meaningful groups.

Imagine a fiction where every normal person rate 2 and every NTR enjoyer rate 4. Then you can present the ratings like this : there are two reader groups that rate this work 2 and 4. That way a reader who self-identify with one group or another can make informed decision on whether to read or not. In reality, you never find such clear cut population.

My problem then revolves around sorting the reader population into groups. For that, I suggest using a machine learning approach.

Now I'm not too sure on the detail, but maybe you can train a neural network "A" to guess the rating a user will give on book "A" by inputting all past ratings for every other book available on SH.

The neural network reduces the high-dimensionality feature space (ratings on every book available) to a lower dimension latent space (reader's profile taste).

In the trivial NTR example, the latent space should show two lumps of users: one for the NTR-enjoyers and another for the normal readers based on their past ratings. That way, book "A" will receive two separate ratings.

This approach can organically create as many groups of readers as needed.

Yes, it's a waste of electricity.

 

[beast_regards]

Let's imagine a series with 4.8/5 average rating. Someone feels it deserves 4/5. In your ideal world (in SH's current rating system), it should be easy, they should just rate it 4/5, according to their personal rating. What if they think average ratings is more important than their own experience? What should they do then, to make the series fall into 4/5 faster?

That's right, by picking 1/5. In an ideal world, 1/5 represents the absolute shit and the shittiest. But this reader doesn't feel that way, for them average rating is the best representation, and they just dislike the series for being overrated. And 1/5 rating is the fastest way to decline it.

That's the core of the problem.

 

[Tyranomaster]

This is why we would use ai to generate memes based on your story. Then readers can judge based on the quality of the memes.

Isn't that just RR advertising? I think we had a thread laughing at some of them recently...

In the 5-star-rating system, however, the writers are under much larger pressure to reach certain value (and once again, math is on the site of the opposition) while under like-dislike system, there is no such thing.

Instead of going on the lowest possible rating, they would just hit dislike.

(1+5+5)/3=3.6
vs.
Negative+positive+positive=mostly positive

Actually, it may lessen the pressure instead.

Then you know there are five people who liked your story, and five disliked: You will clearly see the views are somewhat balanced and the five haters didn't take that much from you. After all, it should look achievable to get one more like.

However, if you see your story is rated 3.0 and your story needs to be 4.5 you would feel it a more of the struggle to get it there, especially after the other values join in, like twos and fours.

There is a valid point in here about 1-star ratings having undue influence when someone is vindictive or just outright dislikes the story. I can see that.

I also agree that +1/-1 system is probably more "just" to the majority of authors. I don't think the rating system is anything more than it is though, and the differences are marginal, and it's not meant to satisfy authors, it's meant to satisfy readers, and to some degree vindictive readers are readers.

Manipulations happen in either system, easily.

The only way to "simultaneously improve the subjective rating that both authors and readers" give to a rating system is to convolute it heavily. My earlier posts try to point that out. I think a better system possibly exists and is doable. It's complicated and requires lots of fine tuning, and will still alienate some people. Twitter/Youtube algorithms do this. The majority of people probably like it more than a random system, but it actively adjusts perceptions hurting a minority. You only get to choose the hurt minority.

Right now the hurt minority is a subset of authors. The hurt minority in a +1/-1 is a slightly different subset of authors, and a larger subset of readers. In a convoluted system, the hurt minority likely shifts from month to month based on tuning parameters.

I advocate a completely different approach:

The goal is not to deduce a single truescore, but separate ratings into meaningful groups.

Imagine a fiction where every normal person rate 2 and every NTR enjoyer rate 4. Then you can present the ratings like this : there are two reader groups that rate this work 2 and 4. That way a reader who self-identify with one group or another can make informed decision on whether to read or not. In reality, you never find such clear cut population.

My problem then revolves around sorting the reader population into groups. For that, I suggest using a machine learning approach.

Now I'm not too sure on the detail, but maybe you can train a neural network "A" to guess the rating a user will give on book "A" by inputting all past ratings for every other book available on SH.

The neural network reduces the high-dimensionality feature space (ratings on every book available) to a lower dimension latent space (reader's profile taste).

In the trivial NTR example, the latent space should show two lumps of users: one for the NTR-enjoyers and another for the normal readers based on their past ratings. That way, book "A" will receive two separate ratings.

This approach can organically create as many groups of readers as needed.

Yes, it's a waste of electricity.

I don't think the problem is actually complicated enough to need a neural network. I think you just have a mousover function that informs people that "Truescore is genre adjusted" as such, the story is roughly rated compared to comparable stories.

Yes, that means it really doesn't have that much meaning because stories that overlap entirely in genres are rare, it's probably closer to accurate for people searching a tag than a straight rating system though. If stories only had like 2 tags, maybe a split system would work better, but there are so many tags you always have to sort of approximate things.

---

Edit: On an added note, we're all talking like steam is the only +1/-1 system, but reddit is a cesspool, and it also uses that system, and is probably a closer approximation to SH than Steam is, so I don't think it's clearcut that +1/-1 is a good system.

 

[beast_regards]

There is a valid point in here about 1-star ratings having undue influence when someone is vindictive or just outright dislikes the story. I can see that.

I also agree that +1/-1 system is probably more "just" to the majority of authors. I don't think the rating system is anything more than it is though, and the differences are marginal, and it's not meant to satisfy authors, it's meant to satisfy readers, and to some degree vindictive readers are readers.

Manipulations happen in either system, easily.

The only way to "simultaneously improve the subjective rating that both authors and readers" give to a rating system is to convolute it heavily. My earlier posts try to point that out. I think a better system possibly exists and is doable. It's complicated and requires lots of fine tuning, and will still alienate some people. Twitter/Youtube algorithms do this. The majority of people probably like it more than a random system, but it actively adjusts perceptions hurting a minority. You only get to choose the hurt minority.

Right now the hurt minority is a subset of authors. The hurt minority in a +1/-1 is a slightly different subset of authors, and a larger subset of readers. In a convoluted system, the hurt minority likely shifts from month to month based on tuning parameters.

You, as the writer, don't want the story being high-rated.

You want your story being found.

And supposedly liked by someone. Not liked as a thumb up, liked as in there are readers who enjoy the story.

The necessity to have a good rating adds the unnecessary complexity to the simple goal.

Being read, or having a good story.

It's no longer about the subjective or objective quality, it is about the additional level of complexity of achieving the goal of being read, one which doesn't revolve around writing itself.

Marketing is a skill, and it is a unique skill that differs greatly from creative writing.

Now, with rating, the writers could be easily hampered by the things which are simply beyond their control. Marketing.

That's the reason professional authors don't interact with anyone outside the controlled environments set up by their publishers.

Because the publishers don't want the writer to be the better marketer, but the better writer. Having both skills is rare. Publishers handle marketing, but can't write...

However, amateurs are forced to handle both the writing and the marketing equally well, while the result is outside of the control, as the rating is ultimately crowd sourced to people not invested in their success. In fact, sales and profit are not their motive.

And more important.

You could improve as the writer.

You couldn't improve from the rating which will remain even after you fixed the error that caused it.

 

[Tyranomaster]

You, as the writer, don't want the story being high-rated.

You want your story being found.

And supposedly liked by someone. Not liked as a thumb up, liked as in there are readers who enjoy the story.

The necessity to have a good rating adds the unnecessary complexity to the simple goal.

Being read, or having a good story.

It's no longer about the subjective or objective quality, it is about the additional level of complexity of achieving the goal of being read, one which doesn't revolve around writing itself.

Marketing is a skill, and it is a unique skill that differs greatly from creative writing.

Now, with rating, the writers could be easily hampered by the things which are simply beyond their control. Marketing.

That's the reason professional authors don't interact with anyone outside the controlled environments set up by their publishers.

Because the publishers don't want the writer to be the better marketer, but the better writer. Having both skills is rare. Publishers handle marketing, but can't write...

However, amateurs are forced to handle both the writing and the marketing equally well, while the result is outside of the control, as the rating is ultimately crowd sourced to people not invested in their success. In fact, sales and profit are not their motive.

And more important.

You could improve as the writer.

You couldn't improve from the rating which will remain even after you fixed the error that caused it.

I get where you're coming from, but I disagree on a few fundamentals.

First, the rating system isn't meant to be for authors. Yes, authors want readers, but authors have no rights to getting readers. It's a hoop to jump through, but there will always be a hoop to jump through regardless of the rating system or lack thereof. You only get to choose the hoop.

Second, authors are not shackled by their old ratings when they fix the error (and no rating system changes that). If they truly fixed enough errors, they should be re-releasing the story as a rewrite. Making small fixes in grammar as they go does not qualify for that. Yes, it takes time. If it didn't, then ratings in general would have less meaning, making the site far less useable for readers (the target audience).

Third, as far as I'm aware (and I've held three major job titles in my life so far, and numerous hats in other groups of local and group administration), every job requires you to be competent in multiple areas that aren't your expertise. You don't need to be a genius marketer, but you need to be competent at it as an amateur writer to succeed long term.

Every professional writer has stories of how they started out by reaching out to dozens of publishing companies and going to writing conventions trying to talk to scouts. That's marketing. They did it too. As an engineer I had to deal with tons of administrative and accounting work. It's not "Outside of their control" it's in their control, and it's reality. Reality doesn't care what we wish was the case. Everyone's brain wishes they could simply succeed with zero effort at everything. The brain is lazy and doesn't like to work. It still has to.

Those who understand the legwork necessary and commit to it succeed more than those who don't. The rating system has no weight on those factors. There are always administrative roadblocks to success, you can change out the roadblock, but it'll still be there to jump through. Fix one, and a different one comes up that people will stumble over to fill it's place.

Change to a +1/-1 system, and small name writers get advantages based on the number of friends they have who can uprate their story early more than it would in a 1-5 rating system (because people are suspicious of stories with exceptionally high ratings with very few chapters). Meaning it turns into a bit of a popularity game (which happened on reddit, and we've seen that result 10 years later). I think a +1/-1 system would lead to a handful of exceedingly popular genres and/or genres prone to overly positive feedback pushing out everything else from trending because the trending system would have less data to go off of. Sure, you could fine tune it, but then you're already having to make the system more complicated.

 

[unlaumy]

A bell curve would be a flawed model because in reality, the points of the bell curve are subjective and we cannot assume ceteris parabus

Yeah you're right. I was too deep on giving justice for the NTR enjoyers. Implementing bell curve would actually harm more people.

That's the core of the problem.

I went overboard explaining the obvious then.

It's a problem that will stay as long as average rating is publicly shown. No matter how you changes the system, these raters will stay, including every suggested systems in this thread. That's what I'm trying to say.

I advocate a completely different approach:

The goal is not to deduce a single truescore, but separate ratings into meaningful groups.

Imagine a fiction where every normal person rate 2 and every NTR enjoyer rate 4. Then you can present the ratings like this : there are two reader groups that rate this work 2 and 4. That way a reader who self-identify with one group or another can make informed decision on whether to read or not. In reality, you never find such clear cut population.

My problem then revolves around sorting the reader population into groups. For that, I suggest using a machine learning approach.

Now I'm not too sure on the detail, but maybe you can train a neural network "A" to guess the rating a user will give on book "A" by inputting all past ratings for every other book available on SH.

The neural network reduces the high-dimensionality feature space (ratings on every book available) to a lower dimension latent space (reader's profile taste).

In the trivial NTR example, the latent space should show two lumps of users: one for the NTR-enjoyers and another for the normal readers based on their past ratings. That way, book "A" will receive two separate ratings.

This approach can organically create as many groups of readers as needed.

Yes, it's a waste of electricity.

I don't think the problem is actually complicated enough to need a neural network. I think you just have a mousover function that informs people that "Truescore is genre adjusted" as such, the story is roughly rated compared to comparable stories.

Yes, that means it really doesn't have that much meaning because stories that overlap entirely in genres are rare, it's probably closer to accurate for people searching a tag than a straight rating system though. If stories only had like 2 tags, maybe a split system would work better, but there are so many tags you always have to sort of approximate things.

---

Edit: On an added note, we're all talking like steam is the only +1/-1 system, but reddit is a cesspool, and it also uses that system, and is probably a closer approximation to SH than Steam is, so I don't think it's clearcut that +1/-1 is a good system.

Ironically enough, because there are divisive views on certain tags/genres, making the ratings suffer from it, the need for a fair rating system itself becomes void. That is to say, the implementation of weighted/group segregated rating fuctions for unpopular tags/genres are no longer necessary, because they're now acting like fandoms in AO3.

 

[RiceballWasTaken]

harm more people

okay now I support the bell curve.

 

[Deleted member 84247]

I think there are better rating systems. At the same time, I don't care. I got a bad review from an alt because of a forum post I made. You just have to accept that some people are stupid. At some point, you can't let trolls dictate how you feel. The internet is full of trolls, and if I let all of them get to me forever, I'd delete my accounts permanently.

 

[owotrucked]

Ironically enough, because there are divisive views on certain tags/genres, making the ratings suffer from it, the need for a fair rating system itself becomes void. That is to say, the implementation of weighted/group segregated rating fuctions for unpopular tags/genres are unnecessary, because they're now acting like fandoms in AO3.

The machine learning suggestion doesn't use tags/genre.

Theoretically it takes the entirety of available rated books as base for the feature space to separate readers organically.

But the neural network needs to be trained from scratch for every single book where you want to split the total rating into multiple ratings.

It's as elegant and energy efficient as going to groceries with a single-use rocket spaceship

 

[Valmond]

I think there are better rating systems. At the same time, I don't care. I got a bad review from an alt because of a forum post I made. You just have to accept that some people are stupid. At some point, you can't let trolls dictate how you feel. The internet is full of trolls, and if I let all of them get to me forever, I'd delete my accounts permanently.

 

[georgelee5786]

5 star system but you should be required to write a review of at least 100 words.

 

[3guanoff]

I do not care about stars or thumbs. Do whatever you would like. However, please do not require me to write 100+ word reviews.
As a reader, I usually dislike reading them as much as I dislike leaving them.

As for genre bias, I do not see the problem.
Please excuse my poor math (if applicable). I never went to high school.

My understanding of statistics(?) leads me to believe that certain genres receiving worse ratings is not a problem.

Here's my possibly flawed reasoning:

If a genre is disliked, all novels with that tag will receive worse ratings. Thus, if a reader searches for a novel in that genre, he will not be able to exclude novels with bad ratings... since most if not all novels in that genre will have bad ratings.

This is like searching for a place that serves burgers in my home country. The burger will be either overpriced, awful, or overpriced and awful. If you want to eat a burger, you know as much but want one anyway. Maybe you like that awful taste or maybe you are a foreigner who only eats burger. In either case, you will not compare burger restaurant ratings with other restaurant ratings.

What about readers who are not searching for novels within that genre/with that tag? Well, if a genre/tag is divisive, that means quite a few people dislike it.
Ratings are meant to help people find things they like. Hence, is it wrong that a disliked tag receives worse ratings?

Those people looking for restaurants in my home country will not wish to find burger restaurants. Why? Only foreigners or people specifically hunting for burger want to eat there. The average fancy person looking to dine out will want to find a place that fits the average taste.

They want their Fantasy LitRPG adventure and you know it.