Rating [1]

The poll....have you seen the number of people who have voted on the poll. And I argue against the 10 point system simply because I hate 10 points. I really do. A 1-5 with half stars maybe. But 1-10 is just incredibly aesthetically ugly in my opinion. Also there are plenty of other aspects of NUF that I find highly more compelling and important. *shrugs*

And just because a small vocal minority complains, does not make it "progress." Goodreads, quite possibly the website with the greatest amount of reviews and ratings for literature on the web uses a simple 1-5 star rating still, after having been around for over a decade tells me one simple fact: More people prefer the simplicity of a 1-5 star rating system. If many of its users had a problem, I think they would have made some changes by now.

 

I don't care if its 10 points or lets us vote 4.5. The end result to me is the same in terms of flexibility. Implementing 4.5 stars should not be difficult as most of it is already there.

But it is progress...if 4.5 is added, you can still vote 4 or 5. Just gives more options.

The 1-5 scale works well when trying to gauge satisfaction, hence why it is commonly used. But when talking about actual rating things, rating on 10 points is more accurate.

PS again when I say 10 points, thats either adding 4.5 or up to 10, I don't care which. Both would work well.

 

It isn't broken and I don't think a 10 point system would be an improvement. If people thought that it was a problem, this issue would have been brought up at some stage in the past. I think the logic is pretty easy to follow.

If you would like to point me in the direction of people on the forums 'complaining for years' about the current rating system on NU, I'd like to see it.

No, it doesn't. Just because people prefer X to Y doesn't mean that people think Y is an issue.

I think that the concentrated information in a 1-5 system is easier for me to read to get a very general idea of what people think about a novel. Spreading that out over a 1-10 system doesn't help that, so yes, I can see a negative aspect towards it. 'End of the day,' we have different opinions about the rating system and that is fine, but saying stuff like 'half of the people in the poll think it's an issue' and 'people have been complaining for years' is a bit of an unfounded stretch.

If anything, the poll results you are pointing to are reason enough to say that the system shouldn't be changed.

 

its bee asked before:
http://forum.novelupdates.com/threads/rating-system.24945/

But it is an issue to some people.

Again, 1-10 or add 4.5, either is fine. If 1.5, 2.5, 3.5 and 4.5 are added it would make no difference. But it would make a lot of difference for people who care about it. 0 negative aspects while giving more accurate ratings.

I am surprised you are not petitioning to remove the decimals at averaging, and not asking for it to be rounded. Since decimals are somehow complex?

The poll shows there are plenty of people who care. If it was rephrased to say add 4.5 instead of 10 points, it would be even larger.

 

Ok, but one thread with 2 people saying they would like the system changed doesn't really make for an argument. If anything, it is an argument that the current system is better because more people voted for that.

It's not possible to please everyone. I would rather a 5 point system, so maybe a 10 point system could be an issue for me. It's not really, but just as a 5 point system could be an issue for some people, a 10 point system could be an issue for others. Neither opinion is more valid than the other.

You keep saying zero negative aspects but I have already said twice that I think the concentrated information is easier to read at a glance. It's just a preference and yours is obviously different but that doesn't mean that my opinion isn't as valid as yours.

You can spare me the condescending remarks. I've already stated my point and you just keep saying the same thing. It has nothing to do with decimals or rounding and if you actually bothered to read what I have written twice already, you would understand that.

The does not show that there are plenty of people who care. It shows that so far, 3 people think that a 10 point system would be an improvement. If anything, it suggests that the system shouldn't be changed because twice as many people have said that the system shouldn't be changed. Stating that rephrasing the question would change the vote is just speculation.

 

rating is worthless, its better for NU abolish rating system.

 

You know what would actually be useful to improve the rating system?.... the ability to remove your own rating.... or you know a confirm rating button..... cuz my darn meat fingers keep accidentally rating things while scrolling on mobile..... pretty much everything I've rated was done accidentally and most certainly does not reflect my thoughts on a novel at all.....

 

You seem to be on purpose trying to ignore the main point I am making. Probably because you feel like you are losing the argument,

Again, how does it matter to you if you can rate 3.5, 4.5 and etc? You end up reading the average anyways in decimal and you can still vote your 3,4, and 5. or w\e. It does not harm you one bit. But to those who want to give more accurate ratings, it matters a lot.

 

I said already, because the results of a 1-5 system are easier to read at a glance. The system isn't there for complete accuracy, it's for a general idea. Losing the argument? This isn't an argument, I've made my point and said why I don't agree with yours, now we are going in circles. The poll results which you pointed to before yet now conveniently ignore are evidence enough that the system isn't needed.

ed. I don't read the average rating, I look at the bar chart.

 

it's useful for someone who have no idea about a title~ sure it have flaw~ still better than no rating at all~ too lazy to read review thing~

 

No the fourth suggestion is necessary for my fifth suggestion — suggested novels.

Suggested novels works like this:

If two people (X and Y) have ratings that are highly correlated across novels they both rated, then if X for example rates a Novel J that Y has not read 5.0, then J wouldn't show up under Y's suggested novels list.

This is one benefit.

Another benefit:
If all ratings are public, then people would be less willing to dish out troll ratings.

Bayesian Rating benefits:

1. Sample size becomes irrelevant — it is irrelevant how many people rate a novel as the Bayesian rating would reflect an adjusted rating based on the sample size. Bayesian Rating algorithm allows you to adjust for limitations in sample size.
2. Bayesian rating adjusts the rating for over rating and underrating.
3. Bayesian rating estimates what the rating would be if everyone in the community rated the novel.

These are the benefits of a Bayesian rating algorithm — to explain why it does all of this would involve me getting I to the technical details of a Bayesian algorithm. NUF doesn't permit LaTeX, so I'll need to upload a PDF describing the properties of a Bayesian Rating Algorithm. I'm currently working on designing one as we speak — others exist, but I don't want to rip them off.

I am not throwing a fancy word about.

 

a 1 - 5 stars system with half star ratings IS a 10 point system.

A 10 point system does not imply numbers 1 - 10 — it simply means that there are ten different points of rating.

This was in the OP, but if it was not clear, I'll go edit to clarify.

 

I doubt a suggested novel system is worth implementing considering that there are already other avenues for people to get recs. I also doubt that people would be less willing to dish out troll ratings because generally the people who give a novel a 1 because there is something they didn't like either feel like their rating is justified or just don't care about your opinion. Showing everyone's ratings publicly does nothing but invite hate.

 

Hmmm, well the suggested novels system. can still be implemented implicitly — the system would still keep track of the rating correlations between two individuals, but won't display it publicly.

Suggested novels would still be an improvement because not everyone uses the forums.

I think the suggested novels feature would be really cool.

 

There is already a suggested novels feature... Why would ratings need to influence on that, the suggestions are already there, and it's pretty easy to find them.

 

TL;DR
None to minimum benefit at best. Requires work to be done. Potential backlash.
Would not advocate, personally...

If anything, we should help users get more acquainted with the biases in the ratings so they can get a good interpretation of the ratings.

No matter what system is implemented, users have to learn about how it works to actually understand what the ratings mean anyways. What would you expect from a 4.8 vs. 4 star xianxia novel? 4.8 for xianxia probably makes it phenomenal considering how many people would give 1* rating for weird reasons.

_________________________________________________________________________________
The problem seems to be that the current ratings is not easy to interpret, particularly for new users of the site. Right?
This arises from biases and tendencies in ratings. For example, the tendency for people to vote in the extremes.

In that case, changing from 5point to 10point system wouldn't really do much.
The biases that exist will still exist.

There's also going to be a difficulty in implementing this, as old ratings will not be adjusted.
This would add yet another bias in the novels that have a lot of ratings currently on NU.
I think it's trivial for everyone to understand that wiping ratings if a new system is implemented will also cause inconveniences.

A 10 point system theoretically is much more accurate. However, this is practically useless.
I won't bore anyone with large spreadsheets, so let's just take one example.
ISSTH ratings on NU vs. on MU
Average on NU is 4.5/5, on MU it's 8.8/10
Very close in results. There's practically no difference in how a user would interpret these numbers.

The theoretical inaccuracies disappear as we have more ratings.
Say a novel's true rating is 4.5
Simply, about half of people who think it should be 4.5 will eventually pick 5* and the other half will pick 4*.
Even with variations, it wouldn't vary much.

In the end, a 10point system solves nothing, and just adds work and potentially problems.

Bayesian average will only smooth out ratings, and only be relevant when sample size is small. It does nothing to fix the biases in people's ratings. It simply adds another bias which smooths out everything by making everything tend towards this predetermined value. So this also isn't a real solution...
One should just realize the ratings mean nothing when there's only a few votes.

_________________________________________________________________________________
What exactly is the problem with these biases existing?
It's in interpreting them.
The real issue is not on getting an objective rating based on certain criteria. That's not too useful.
What is useful is a rating system that can very roughly allow one to estimate the quality of a novel while sifting through the long list of stuff to read.

How do veterans do it? We have an instinctive grasp of what the average scores are. Through going through many ratings, we have a rough idea of what certain scores mean. We also have a good grasp of what kind of tropes are well-liked by the community.
For xianxia, around 4* is barely decent quality. Something around 3.5* is probably 90% cliches, 10% original content kind of deal.
Every one of us has a different cutoff for different things, too...

This isn't something that can be solved, as a rating system based on just simple numbers is very limited.

That's why we have on top of that genres, tags, and reviews. Even then, the better you know about the biases that exist in the community, the more you can get out of them.

Maybe instead of trying to fix a system that isn't broken, we could implement a system that complements it.
For example, showing how a novel's rating compares to others of the same genre. This has the added benefit of not having to remove any old systems, meaning no grandfathered ratings.
With respect to having a 10-point system or Bayesian average, it'll probably be more useful to just show average rating of, say the top 100 most read novels of a genre.

 

Someone remind me why I forgot to link this post in the first place? That's usually the first thing I do when someone makes a thread about an issue that has already been brought up before...

In any case, here is the standing by Tony on the issue of ratings. Check out the actual thread for the context.

 

The 'suggested novels' proposing is something entirely different.

I don't know what's in place, but it doesn't do what the feature I'm proposing does.

It is embarrassingly easy to adjust all old ratings to accommodate whatever new point system were going with — just multiply all ratings by c where c is (n/m)
n is the number of points in the new system.
m is the number of points in the old system.

This is simply untrue — most likely the vast majority of people who thinksl it's 4.5 will rate it 5 star or not rate.

A 10 point system does not destroy any ratings — simply multiply all extant ratings by two.

Bayesian Average does A LOT for novels — but this isn't the thread to solicit the implementation of a Bayesian algorithm.

 

No. If I had wanted to give 4.5 to ISSTH, but my current rating is 5. How do you adjust that? This would give a bias to the ratings since the old system and the new system are different.

Do I really have to give an in-depth explanation as to why introducing a bias unnecessarily is not a positive thing?

How do you support this claim, exactly?

Intuitively, if someone's rating falls between 4 and 5 and true rating is 4.5, about half would be below 4.5, thus being closer to 4 and half would be higher than 4.5 thus being closer to 5.
In the case where it falls between 4 and 5 and they're indifferent, they would pick at random. Some may tend to make conservative ratings while some may say even though it has imperfections, it deserves a 5 anyway.

Comparing the average on NU and MU, we see that this is indeed the case. The average isn't far different between the 5 and 10 point system, which suggests the split is quite even, such that the 5point system approximates the true rating just as well as the 10point system.

I didn't say it destroys things. I said it provides no practical benefits, while adding work that needs to be done. And there is potentially a lot of people who will want things to go back to the old system...

As I said, if you simplify things that way, you will make it so no votes are on 1,3,5,7,9 among the old votes.
I did say it won't matter, but if you want to convince Tony to implement a 10 point system, I'm assuming you think it does matter. If it does matter, then this is a huge flaw that is introduced in the ratings, which can only be fixed by making every reconsider their ratings. Not very practical.

If you just want to introduce the 10 point system without caring about all the imperfections that come with it (unnecessary imperfections, mind you). Then you're not gonna be convincing anyone.

No, it really doesn't... You can make a thread for it and I'll reply there if you want.
Although I think it doesn't really do anything, it's much more likely to be accepted than the 10pt system, I think.

But very briefly, all it does is add in a bias which cannot be justified.
It would simply amplify what I personally think is a problem with ratings... Rating based on genre, and not on content.

I highly suggest checking out the feature before you comment on it's functionality.

But in short, everyone can give recommendations for similar novels just like they give ratings. You're not limited to one suggestion. There's a number on the side that tells how many people gave the same suggestion.

What you mean to suggest is something like Amazon where they show you "other people who bought this product also bought..." right?
This feature is exactly the same, except it's not automated. And I actually mean literally; not just somewhat. It's literally, "I read this, and I also liked this other novel". It actually does more since we are consciously making a decision to recommend something similar, not just something that's just generally to our liking.

Automation requires some complex algorithms to be as accurate. And if it's not accurate, it kinda beats the point... There's also things to consider such as people not rating everything they read, or how since most people tend to rate 1/4/5, everyone's ratings will probably be highly correlated.

 

You manually go and adjust your current rating — no bias is introduced.

I've seen this feature; it is nothing like what I proposed — which you would know if you read my proposal.

No. Look at my post.

It matches your tastes in novels.

Other people with similar tastes as yours also like such novel.

What I'm proposing is not currently implemented on any system I know of, and is an original idea of mine.

I can tell you that this is farthest from the truth. Everyone's ratings aren't going to be highly correlated — I'm willing to bet my life on this.

As a statistician, I can tell you that the probability of this is of the same order of magnitude as the probability of the world ending tomorrow.

You don't seem to like my ideas — fair enough — your "refutations" aren't legitimate concerns though. They're not valid refutations — now this doesn't mean that my ideas should be implemented, but you haven't raised legitimate reasons for why they shouldn't.