Some tag voting changes

Posted in

#26 by rampaa
2020-06-20 at 14:45
< report >I think the power of negative tag vote should be changed. Consider the following:

-3 vs +1 => -1
-3 vs +2 => -0.5
-3 vs +3 => 0 (mind you, tags with average score of 0 is being treated as "downvoted" tags by VNDB)

So no matter the case, when someone disagrees with a positive vote, they always have the upper hand. Sure, when more than 2 people vote on a tag this becomes less problematic but there are lots of cases where you won't get 3 votes on a tag.

I propose changing the negative vote strength to -2, that way the following will happen.

-2 vs +1 => -0.5
-2 vs +2 => 0
-2 vs +3 => 0.5

Which seems way more reasonable to me.Last modified on 2020-06-20 at 15:09
#27 by schlaefer
2020-09-02 at 19:48
< report >@26 I think that would only defer the problem.
If a tag applies, but not majorily you vote +1 or +2.
Negative votes are for removing tags that don't apply at all, not for tweaking the value. (from my understanding). So it shouldn't make a difference what the value of the positive vote was.

My suggestion would be to compare the number of upvotes with the number of downvotes and hide the tag if the latter is larger (maybe multiplied with a constant before). If the tag is shown than with a score equal the average of all positive votes.

So, for 3 votes with +3,+3,-3 you could still get a 3.0, instead of 1.0, which would change the significance of the tag.
And it would prevent removing three +1 votes with a single -3.
#28 by rampaa
2020-09-02 at 22:46
< report >@27 If two/three people vote with +1 but only a single person disagrees then we probably shouldn't say that that tag does not apply to the game at all. Saying that might even encourage people to vote with greater tag scores just to make it harder to remove the said tag, even if it's not warranted. Your suggestion would solve this problem quite nicely.

However, consider the following: Let's say a considerable number of people can't agree whether or not a certain tag applies to a certain VN due to whatever reason and suppose that number of people who think the tag applies is only slightly more than those who don't. In that case, if we were to only took the average of positive scores, we wouldn't see the effect of the opposition at all, and I am not sure if that's a desirable outcome. I feel like the score of "10 +3 votes vs 9 negative votes" shouldn't be equal to that of "10 +3 votes vs 0 negative votes". If there's a strong opposition then I think the displayed score should reflect that one way or another.

So here's another idea: Negative votes shouldn't have a static vote score but they should change the displayed score nonetheless. See the following:

x: number of positive votes
y: number of negative votes
z: average of positive votes

(The following assumes that there's at least one person who voted on the tag. It also assumes that in case there's no positive vote, "z" is initialized with a positive number.)

if(x!=y) displayed_score = z*(x-y)/(x+y)
else displayed_score = 0.5

This would prevent the problems like removing three +1 votes with a single negative vote and it would still allow the negative votes to affect the displayed score (which, IMO, is a benefit).Last modified on 2020-09-02 at 23:45
#29 by schlaefer
2020-09-03 at 17:06
< report >@28 You are absolutely right. I noticed that problem but wasn't sure if it would actually be an issue. But you are right, when the very tag definition isn't clear to people that could happen.
Like when you have to decide if hair is brown or blond. Everyone could agree it's either of both. But if the actual color is something between it becomes hard to settle on one.

I think your formula goes in the right direction. But I don't understand why you would initialize z. If there is no positive vote the score should be just 0.0 and everything equal or less than zero should not be displayed.

Also I feel like the negative votes change the score too much. I would slightly change it to
displayed_score = z*(x-y)/x
Removing the last y in the divisor. For +3,+3,-3 that would result in a score of 1.5, which seems to me more fitting than 1.0. And for 10x +3 vs 9x -3 it would be 0.3 instead of 0.16. Though at this point both scores are very small anyways.
My point is, if a number of people thinks a tag applies and half as much think it does not, than the score should be halfed.
#30 by rampaa
2020-09-03 at 18:13
< report >
when the very tag definition isn't clear to people that could happen
I think it can happen even if the tag definition is very clear. Because the story might not be clear and people might have different interpretations because of that.

But I don't understand why you would initialize z. If there is no positive vote the score should be just 0.0
Because I sometimes down vote tags even if there's no positive vote for it. There are, in my opinion, some instances where this makes sense. Take link for example. At first I thought the game lacked the functionality but afterwards I found out that it did actually have it. So I've down voted the tag to convey that the game had the functionality. Admittedly this more due to absence of a complement tag, if we had a tag for "Keep Voice on Click Option" as well, this wouldn't be necessary. But I barely persuaded Warfoki to make No "Keep Voice on Click" Option a thing.

Anyway, I digress. I would prefer z to be initialized with 3 in the case of there being no positive vote, because I think it would make the intend more clear and impactful. But I guess it doesn't matter that much either way.

And far more importantly, there's a good chance that we are discussing this for naught, because Yorhel doesn't seem to be interested at all. ._.
#31 by schlaefer
2020-09-03 at 20:48
< report >I see now. I was talking about the main page of a VN where the score determines the order and font-size and everything below 0.0 isn't displayed anyways. But for the modify tags page it does make sense.
I think if there are more negative votes than positive votes it would be viable to only use the number of votes and either use the difference or ratio with a cap of -3 to calculate the score.

I guess we had some good ideas, but it's also not really that important. So we can leave it at that.


You must be logged in to reply to this thread.