Mod Review Score is not calculated correctly.

RiddleTime · Nov 17, 2022

As I have a mod with 16 x 5 stars and 1 x 4 star. That averages (84/17) out at 4,941.
The review star system shows it's got a 4.5 star rating, I find this to be not correct and misleading. If the actual average score would be 4.6 or 4.7 this would be reasonable, but at the moment it makes it look like half of the reviews are 4 stars and half of them are 5 stars.

At what average would the "stars" go back to 5?

I think the system for "stars" should reflect the actual average and not floor out immediately to a 0.5.

And yes you can hover the 'rating' but the website UX is not providing you this knowledge.

Neilski · Nov 17, 2022

Some facts about the rating system, some of which are not as you might expect:

the displayed stars and half-stars reflect a score which is rounded down the nearest half; thus only a score of exactly 5 will show all 5 stars
two averages are calculated - a weighted average which is used when you use the filtering to sort by rating, and the true average which is used for the stars and is displayed when you hover
the weighted average is essentially used to prevent sorting a mod with a single 5-star rating above a mod with 4.9 average and hundreds of ratings (the calculation effectively pretends that every mod has 10 hidden 3-star ratings); this has no effect on the displayed stars or value shown on hover

It might make sense to also display the average rating in plain text...

RiddleTime · Nov 17, 2022

It would good in my opinion to sort based on the true average.

This means that even with 99 five star ratings and 1 four star rating the displayed stars would still result in 4.5, right?

Neilski · Nov 17, 2022

RiddleTime said:
This means that even with 99 five star ratings and 1 four star rating the displayed stars would still result in 4.5, right?

Yup, 'fraid so.

RiddleTime said:
It would good in my opinion to sort based on the true average.

Well, would you really want a brand-new mod with a 5-star rating to sit at the head of the list, beating the mods with hundreds of 5-star ratings and a single 4-star?

RiddleTime · Nov 17, 2022

Neilski said:
Yup, 'fraid so.

Well, would you really want a brand-new mod with a 5-star rating to sit at the head of the list, beating the mods with hundreds of 5-star ratings and a single 4-star?

Ooh well, it's just a rating, I care more for the actual usage of my mod that makes it worth making.

As for the question, not so much.

But I prefer to at least have fair reviews based on the existing mod and not reviews based on feature requests for a mod which should be placed in the discussion thread of it in my opinion.

Neilski · Nov 17, 2022

RiddleTime said:
But I prefer to at least have fair reviews based on the existing mod and not reviews based on feature requests for a mod which should be placed in the discussion thread of it in my opinion.

There's certainly a grey area at the boundary between legitimate reviews and those which belong in the discussion area, but we do regularly delete reviews which clearly belong in the latter (e.g. a 1-star review with a question about how to make it work). If you see reviews which you believe to be questionable, you can report them.

RiddleTime · Nov 18, 2022

I would have a suggestion for the stars which would be clipping the stars based on the width of the html element instead of using a half star system. It can be done with CSS, if any help is required feel free to message me.

Neilski · Nov 18, 2022

RiddleTime said:
I would have a suggestion for the stars which would be clipping the stars based on the width of the html element instead of using a half star system. It can be done with CSS, if any help is required feel free to message me.

Interesting idea (I'm rubbish at CSS), though I'm not sure that on many screens (especially mobiles) it would be possible to really spot much difference between 4.5 stars and 4.6 stars, or 4.9 and 5.0...
What do you reckon @Bram Hengeveld?

RiddleTime · Nov 19, 2022

CSS wise it can be done using the clip-path property on an element.

An example, how it could be done:

// 75% of 5 star rating
.ratingStars {
    clip-path: inset(0 25% 0 0);
}

Neilski · Nov 19, 2022

RiddleTime said:
CSS wise it can be done using the clip-path property on an element.

An example, how it could be done:
// 75% of 5 star rating .ratingStars { clip-path: inset(0 25% 0 0); }

I'm curious to know how that would look if you just temporarily hack it into the page. However, I have to tell you that changes to the way the site displays star ratings are probably not going to happen.

Ole Marius Myrvold · Nov 19, 2022

Neilski said:
Well, would you really want a brand-new mod with a 5-star rating to sit at the head of the list, beating the mods with hundreds of 5-star ratings and a single 4-star?

I have a suguestion. Queggestion... Suggestion - Question!
Would it not be possible to have some sort of lower number of reviews a mod needs to be counted on the sorted list?
I mean, that might catch out some great niche-mods, but could be a solution? If even possible within our software.

Neilski · Nov 19, 2022

Ole Marius Myrvold said:
I have a suguestion. Queggestion... Suggestion - Question!
Would it not be possible to have some sort of lower number of reviews a mod needs to be counted on the sorted list?
I mean, that might catch out some great niche-mods, but could be a solution? If even possible within our software.

Well, we could consider code tweaks, but the fake/weighted average thing sort-of does a bit of that anyway.
In any case though, my guess is that the most popular sorting is likely to be by the number of downloads, given that ratings are a bit... fickle?

Ole Marius Myrvold · Nov 19, 2022

Neilski said:
Well, we could consider code tweaks, but the fake/weighted average thing sort-of does a bit of that anyway.
In any case though, my guess is that the most popular sorting is likely to be by the number of downloads, given that ratings are a bit... fickle?

Idk. I honestly have never sorted mods on RD, I search if there is something I want

RiddleTime · Nov 22, 2022

Neilski said:
Well, we could consider code tweaks, but the fake/weighted average thing sort-of does a bit of that anyway.
In any case though, my guess is that the most popular sorting is likely to be by the number of downloads, given that ratings are a bit... fickle?

Well if such rating is considering weights.. Is there a pattern in reviews that occurs a lot with download rates based on update interval or maybe other metrics could apply.. :O

Neilski · Nov 22, 2022

RiddleTime said:
Well if such rating is considering weights.. Is there a pattern in reviews that occurs a lot with download rates based on update interval or maybe other metrics could apply.. :O

Not sure what you mean here.

RiddleTime · Nov 22, 2022

Neilski said:
Not sure what you mean here.

I think the actual frequency of downloads straight after a release an indicator in how "popular" a mod is.

Neilski · Nov 22, 2022

RiddleTime said:
I think the actual frequency of downloads straight after a release an indicator in how "popular" a mod is.

Possibly true, but certainly seems too complicated a metric for a ranking system

RiddleTime · Nov 22, 2022

Neilski said:
I'm curious to know how that would look if you just temporarily hack it into the page. However, I have to tell you that changes to the way the site displays star ratings are probably not going to happen.

Not totally accurate since there is a spacing in between the stars (which could be calculated). But it should give an idea as of what it would look like.