Dungeons and Dragons Wiki talk:Rating Articles

From Dungeons and Dragons Wiki
Revision as of 23:33, 19 June 2012 by 108.6.239.14 (talk) (→‎Voting)
Jump to: navigation, search

Old Discussions

Discussions about the RC replacement can be found at the archive. - Tarkisflux Talk 03:38, 4 July 2011 (UTC)

Very Little Information

All that this new system tells a reader, at the moment, is whether or not more than 50% of the raters like the article. That can be an insignificantly small distinction, or a very very large one -- for any useful information at all, the reader needs to read the ratings on the talk page, which begs the question: why bother putting the voting results on the front of the page at all? It would be far more informative if a percentage of favor were displayed, and a number of votes so readers have an idea of the significance of that percentage (something like "33% — 2 of 6 raters liked this article."). --DanielDraco 22:11, 4 July 2011 (UTC)

It actually tells people one of 4 things: whether more people like it than not, whether equal amounts of people like and dislike it, whether more people dislike it than not, and whether the article has enough more people liking it than not to qualify to be a community favorite. It also includes a link to the talk page to look at the ratings themselves, which is something people should do since we're letting pretty much everyone rate and the substance of the votes is as important as the votes themselves IMO (and that is not something we have room for on the main article). So yeah, it is less informative than your suggestion, but I don't see why giving them more information is helpful in this case. Even if we give percentages they still probably need to look at the substance of the ratings, and the percentage itself could be misleading because of the community favorite bit (since the percentage could be between 100 and 50 depending on rater sample size and still qualify an article). It's also not that "2 of 6 people liked this article", it's that 2 people liked it and 4 people disliked it enough to write a negative review. That distinction seems lost in your wording, especially in cases where "0 of 4 people liked this article". That sounds much more neutral to me than the reviews are going to be.
That said, I'm not opposed to changing the display. I'll even put this one up if enough users want it and I get outvoted. I just think that we have enough information to direct people who are interested in the ratings to look into them, and that's about as good as we can get. - Tarkisflux Talk 22:34, 4 July 2011 (UTC)
Well, my concern is the fact that a difference of one person is never statistically significant*, and that that's (for odd numbers, at least) all that stands between the above fifty and below fifty thresholds. Only having a 50% threshold simply does not provide any meaningful information.
*Nevermind the fact that we're not likely to ever have much statistical significance in this, but...may as well maximize it.
Maybe taking a page from Rotten Tomatoes' book and terming it more like "X of Y ratings were favorable" would be more apt. And what's this about community favorites? I don't see anything explaining that on this page. --DanielDraco 23:50, 4 July 2011 (UTC)
If we want to track two numbers (and you're suggesting likes and total here), why not just track likes and dislikes directly? "X users like this, and Y users dislike this" or whatever? It's extra numbers in the author template, but it's not asking people to do subtraction so it's probably a wash. Unfortunately, it's also not something that's easy to bot edit, so someone would need to do these changes manually (which is a mark against it as far as I'm concerned, but something that I'd work with if there was enough support for it).
As for the community favorite thing, that's just what we're calling the featured articles now (though I guess the name is still up for debate since no name ever got real support behind it). It's any article with 4 more likes than dislikes (though we could add a percentage requirement to that as well if we wanted, like 75%), that someone has written up a short blurb on and put into the rotation on the main page (requires minor admin work due to cascading page protection). It's not a supported thing yet (need to figure out a new main page layout since it will displace something, the doc and functionality is easy), and since I wanted to get the ratings up for users yesterday I blew off putting it in the page for the time being. - Tarkisflux Talk 01:24, 5 July 2011 (UTC)

Community Favorites

The section is filled in. Both the name and the 4 more likes than dislikes thing to qualify them were pulled from the previous discussion. If anyone wants those changed, some suggestions would be nice.

The random rotation thing mentioned in the section was talked about a long long time ago, and basically consists of doing two choose lists: 1 with the most recent 9 articles and another 1 with everything else. The main page would then show one of the 9 newest, or one of the older ones. We can adjust those percentages to whatever we want though, in case those numbers look bad for some reason.

Also, here's what I'm thinking of migrating the Main Page to in order to accommodate the favorites: User:Tarkisflux/sandbox/Main Revamp. The favorites would take over the current tribute slot (though the tribute would probably go on rotation), and that should give them a sufficiently sized box to display stuff and be noticed. - Tarkisflux Talk 05:24, 10 July 2011 (UTC)

More recent stuff

Recently people have been talking about us, and some of us here have been talking back, and DD and I talked about the concerns that had been brought up and which ones were valid or not. You can find both the link to the thread and the chat log between us here. --Ghostwheel 21:14, 16 April 2012 (UTC)

Having read the chat log above, I'll give a bit of a tl;dr for those who may not be inclined to read it: I particularly agree with the premise that the initial rating system (i.e. "most people like/dislike this article", etc.) is rather misleading and could be improved upon, both because of the ambiguity of the word "like" and the issues with a +1/-1 grading system. Their proposal of a system of each rater giving a score of between -2 and +2 and then averaging them allows us to use a system that more accurately reflects how strongly we feel about something in one way or another. I think it has promise. - TG Cid 22:28, 16 April 2012 (UTC)
This is a finer detail, but I still think 1 to 5 or 0 to 5 is better than -2 to +2. Firstly, it's a ubiquitous scale; everyone is very accustomed to using it. Secondly, it better reflects what we're actually judging. After all, shittiness is not actually a thing, any more than darkness or cold are things. It's just absence of merit. It's easier to just conceptualize it as a continuum with no qualitative categories than it is to throw in a turning point between "bad" and "good" at zero. --DanielDraco 00:08, 17 April 2012 (UTC)
We've had this discussion before. I am not opposed to the idea, but here are some of the issues that need to be addressed if we retain the same template based system:
  • A simple average is not sufficiently granular for determining CommFav or CommOpp IMO. I would want a sufficiently high/low average as well as a sufficient number of raters.
  • There is no automatic way to determine the average. You are asking each new rater to do math, or to append their score to a list like 2+3+4+5+2, and increment the raters total by 1 (since there is no easy counting code). I think the latter is a better call given my preference for tracking the number of raters anyway.
  • There is no easy migration process for this setup. Is the proposal that we scrap all previous ratings, or is someone volunteering to go through and manually update / delete things?
The above assumes that we're not moving towards a no-template system, like the stars or whatever that appear on other wikis. Anyone proposing we start using one of those needs to take it up with Surgo, as he'd have to install it. - Tarkisflux Talk 01:32, 17 April 2012 (UTC)
0-4 or 1-5 is better than 0-5 since it allows people to choose a point right in the middle if they don't feel good or bad about the material.
Also I mentioned that for the purpose of favored/opposed articles, there would have to be a rating average below X/above Y AND Z number of raters before it could become favored or opposed. And I'd be fine going back and updating my own favors--I think most people don't have a problem with that, it's easy enough. --Ghostwheel 02:09, 17 April 2012 (UTC)
There is a category and a list of pages that a person has rated, broken down by person, specifically for the purpose of getting people to update legacy ratings, and most people still haven't done it. Some of the old raters aren't even active anymore. Any migration scheme needs to address what we do with old ratings, as I do not have much faith in the majority of them being moved over. I would be fine with a direct translation of the really really old ones to 0/2/4 points (instead of 0/1/2), but someone still needs to go through the pages and update all of the author blocks with the new rating values and the new num_raters field. Updating is a non-trivial and not (easily) possible to automate process that needs a roadmap and some serious commitment, or we need to dump everything prior and start fresh. - Tarkisflux Talk 03:40, 17 April 2012 (UTC)

I consider fresh starts (perhaps after a two-week grace period) for ratings to be the only reasonable position to take. We can't exactly ask inactive users what they think of the ratings they previously placed on articles, especially if those articles have changed since then. Frankly, I also believe that if you actually cared about your ratings, you would modify them. So yeah, I'm all for a clean slate approach. - MisterSinister 04:49, 9 May 2012 (UTC)

This is reasonably easy to code for and pretty workable. It's a migration approach I would support, since it would actually get done. - Tarkisflux Talk 04:54, 9 May 2012 (UTC)
I suppose that's a fair idea, given our difficulties in reaching certain people. That said, there are certain legacies that can (and should) be updated, so I think the grace period should also be present to allow that to potentially occur. - TG Cid 04:58, 9 May 2012 (UTC)
I have determined a method to abuse string processing that will give us a total number of raters without having to run a second property. So we only need to list raters (like we already do) and some method of tracking their ratings (like we already do). If anyone wants to write up a new proposal for rating articles that isn't vastly more complicated than the present one, they should do that.
Additionally, I will flat out veto any fresh start migration proposal that does not include a 2+ week transfer period. So time to update legacies will be there, as well as an easy way to determine what needed updating, for any who would take advantage of it. - Tarkisflux Talk 06:05, 9 May 2012 (UTC)
Yeah, really, blank slate is easiest. It's not like we won't have a rating flurry once the system goes in place. Also it looks like more or less everyone is on board with a 5-scale (whether it's -2 to +2 or 1 to 5 is an aesthetic choice and can be decided in the final stages). The big thing that we're still missing to make the proposal complete is thresholds. But it occurs to me that we actually can't determine what a reasonable threshold would be without first getting a good number of pages rated -- what if we set the Fave threshold to 4.0 (with however many raters), but then we get a fuckton of articles in the 4-4.5 range simply because of the way people vote? We need to see where the distribution falls before drawing lines on it. So what do you think of the idea of implementing it for a trial run without thresholds? We'd need a page explaining (as guidelines rather than rules) what we want people to consider in rating, and that's about it. After it's in place for a while, we look at where the numbers fall and get a sense of where we should set the thresholds. We could just use the current list of community favorites in the meantime. --DanielDraco 18:38, 9 May 2012 (UTC)
I'd rather define a metric for the rating (and thus for the CommFavs) and ask people to work from that than try to deal with determining thresholds based on voting patterns that are likely to change over time in the absence of a guideline. Telling people that "a 1 is for things you really hate, a 3 is for things you are pretty neutral on but would probably allow if someone wanted to play it, and a 5 is for things that you love and would run in your games right now" doesn't seem like a stretch. It also allows us to just set the CommFave line at 4.75 average and above (assuming ratings of 1-5) and be done with it. That's just setting it such that, on average, 3/4 of the people who rated it liked it as much as they could and 1/4 liked it just slightly less than that. And that sort of top end community backing seems like the sort of thing we'd want for an actual CommFave. Setting it at 4 or so just means that most people liked, and doesn't really suggest an actual favorite to me in any sort of arrangement where granularity is supposed to matter. If you want to argue that we'd still get a ton of articles in that range anyway or that people aren't going to want to rate on that sort of scale, that seems like an argument for either more granularity (0 to 10 maybe) or a still higher average, not an argument for a potentially flexible requirement. - Tarkisflux Talk 03:43, 10 May 2012 (UTC)
Yeah, I guess that works. And yeah, if a requirement as high as 4.75 yields too many Faves, that would probably be a reason to increase granularity. The greater concern with that threshold, I think, would be too few Faves -- in which case it would need to be lowered. But it's not that big a deal to tweak it later. It was just a thought. --DanielDraco 04:55, 10 May 2012 (UTC)
Tarkis asked me to chime in here. I really, really like the current like/dislike system (though maybe it shouldn't show up in the author template with so few ratings, we don't really have enough people rating constantly to do anything else). I think it's what we should have had from the very beginning. It's simple enough that anyone who shows up can understand it and say they like or dislike something. It gives us a good pool to put articles on the front page. It does exactly what it needs to, and no more. Surgo 03:21, 17 June 2012 (UTC)
Yes, but people generally understand a 1-5 system pretty well too. There are two issues that I see with a binary system. Firstly, it depends on a relatively large number of votes to give any helpful information. We don't get many votes. Secondly, it is more like Rotten Tomatoes, where we might prefer Metacritic -- that is, instead of saying how good most people think it is, all it can tell us is how likely someone is to "like" it to any degree at all. It might be a masterpiece of homebrew, or it might be okay -- the vote will be counted the same, and the users will not get information to distinguish between them. --DanielDraco 03:26, 17 June 2012 (UTC)
I'm not looking to give that much information. I think "Most users like this article" is plenty of information. They can then read the article and see if they agree. A 1-5 system will suffer from the same issue of needing a large number of votes. Surgo 03:27, 17 June 2012 (UTC)
Well, not really. If a user sees that one person likes an article, that tells them next to nothing. They learn the vaguest possible notion of one person's opinion. If a user sees that one person gave an article 5/5, they at least know just how impressed that one person was. It provides a little more information in a way that is equally succinct and equally easy to understand -- I don't see any downside whatsoever to changing to a 1-5 system, and I can certainly see several upsides. --DanielDraco 19:17, 17 June 2012 (UTC)
And also, even if we were to stick with like/dislike, we certainly need to render the information better than "most users like this article" or "most users dislike this article". If the range of people's possible reactions were to be numerically represented in a perfectly accurate manner, it would be a continuum. We're lowering that resolution down to a binary system. Which is alright, since it gives us some different data to work with -- a Rotten Tomatoes system rather than a Metacritic system, to go back to my earlier comparison. But it does end up simplifying the information and losing data fidelity. So then when we take the trends we find in that simplified data and lower that resolution into another binary system, we've lost even more data fidelity. It's worse than giving a measurement with only one significant digit, because even a single-significant-digit datum in a base-10 number system gives FAR more information than we provide. We're doing an unforgivable amount of rounding, and there's no good reason for it. When the statements "50.01% of users thought this article was pretty okay" and "100% of users thought this article was better than sex" are rendered as equivalent outcomes, there are some really big issues at work with how data is being displayed. --DanielDraco 19:27, 17 June 2012 (UTC)

→Reverted indentation to one colon

I'm not opposed to listing actual numbers of likes/dislikes, but I am 100% opposed to ever moving back to a scale. We tried that multiple times, and it never worked. Surgo 19:58, 17 June 2012 (UTC)
Opposed such that you'd vote against it, or opposed such that you'd veto it Surgo?
And since I've figured out how to do more stuff with lists of names, showing totals of people who like, dislike, and neither (which probaly needs another name) an article along with a rating based on the average is pretty easy and would probably simplify the rating process slightly. - Tarkisflux Talk 21:13, 17 June 2012 (UTC)
I'm not Green Dragon, I don't do nonsense bullshit vetos against the community. But I will very strongly speak against it. We have tried it a number of times, even going all the way back to the paleowiki, and every single time it has completely failed at every objective. Surgo 22:55, 17 June 2012 (UTC)
Fair enough.
I'm pretty neutral on a larger scale or a not larger scale, but if people want to display more information or have different qualifications for ratings categories, we can totally do that. Here, I'll even write up a proposal for it:
  • A rater adds their name to one of three parameters within the Author template: |liked by=X, |disliked by=X, or |<whatevered> by=X. There are no longer any numbers to update.
  • These names are turned into numbers and used to determine an average rating. For articles with less than 4 ratings, the total is still divided by 4 so that we get nice numbers for the following ratings categories.
  • Community favorites are articles with an avg rating of 0.8 or above. This would correspond to 4 likes (due to the minimum divisor), 4 likes and a pass, or 9 likes and a dislike. I suspect the latter one will be hard to reach, but it's possible to do. It's also harder to overcome than our current setup.
  • Generally liked articles are those with an avg rating between 0.4 and 0.799. This would correspond to 2 likes, 3 likes and a pass (or a dislike), or 5 likes and 2 dislikes.
  • Mixed or insufficiently reviewed articles would be rated between -0.399 and 0.399. Since generally liked would start at 2, the only time insufficiently reviewed would happen would be with 1 rating only, but it could be split out if we wanted.
  • Generally disliked and Community opposed articles would have negative numbers ranges equivalent to their liked opposites.
  • Numbers of likes, dislikes, and neutral or whatever are shown in the ratings data, along with the general ratings category. Average rating is not.
  • Ratings data is displayed in a more obvious area withing the author template, and changes color with the rating category. This section replaces the Homebrew notification banner.
Aside from visual differences, the only real change in the above is the move to average instead of net for ratings categories. It puts more weight on dissenting views, which I think is probably a feature rather than a bug. It also broadens the central neutral category quite a bit, which is something I want to do anyway.
This would still require a lot of manual updating of rated pages, but would simplify future ratings additions by removing the current totaling step. - Tarkisflux Talk 00:05, 18 June 2012 (UTC)
Regardless of how we do it, one thing that I feel adamant about is the need for more granularity--a 0 to 4 rating system would well. The reason for this is because time and again I find an article that I think is decent but not amazing, and have no way to differentiate from articles that are amazingly good. IMO this is a big problem, and regardless of how we do it, we need more granularity in the rating system. --Ghostwheel 03:58, 18 June 2012 (UTC)

Frankly, I think we need to address the bigger elephant in the room - more than half the damn articles on this wiki are not rated at all. Whatever granularity we do or don't have makes no difference if nobody is fucking using it, which is increasingly what I'm seeing as time goes by. This system only works if everyone rates - and currently, the number of people actually rating anything on this wiki seems to seriously number in the single digits.

Thus, before we screw around with any granularity changes, I'd like to repeat something I proposed to Tarkis a while ago - make a really ugly banner for unrated articles that either goes away or becomes less ugly when these articles are rated. This would incentivize people to actually get ratings for their articles, and if we use those banners to hold rating information, it would both become more visible and make people care about it more.

Thus, until this is addressed in a satisfactory manner, I'm opposed to all changes to the rating system. Full fucking stop. - MisterSinister 04:42, 18 June 2012 (UTC)

Maybe some sort of "credit" system, where you can create something if you've rated 3 articles for every one you create? Not perfect (and is very bad--don't do it, heh) but any other ideas for promoting rating? --Ghostwheel 06:48, 18 June 2012 (UTC)
Ghost has stated that he specifically does not want to rate things without greater granularity. His contention is that with greater granularity there is a lower barrier to liking something, and people (including himself) would rate more. I think.
And I should have put "unrated" and "uncommented" in the above proposal, since I want those as separate rating states on the ratings information box. And I want them to be ugly, the "uncommented" slightly more than the "unrated". - Tarkisflux Talk 07:15, 18 June 2012 (UTC)
(Further comments that were here before MS pulled his) I agree with encouraging ratings, but we need to do it concurrently with any ratings change. If we wait to get a big pile of ratings before considering any changes, there will be sufficient conversion overhead as to make a change extremely difficult. And one of our migration plans is to burn them all down after all. Does anyone really want to get a bunch of ratings and then get rid of them or ask people to do it again? So while I'm off making a table addition to annoy people into rating and commenting on things (tomorrow, it's late), let's carry on with discussions of what to populate it with afterwards. Or not, I really don't care. I just had this conversation a couple of years ago and it didn't go anywhere then either. - Tarkisflux Talk 08:28, 18 June 2012 (UTC)
I saw someone suggesting a Metacritic form of rating... It'd be much, much easier if there was a way to see how people on other popular tabletop gaming forums see the homebrew, then draw that as a form of "critic rating". Of course, it wouldn't work given the idea of the posters on each forum (WotC no longer having the same people as before and mostly focusing on 4e/Essentials/5e, Paizo oogling it to Pathfinder, GITP being just too darn nice, Min-Max Boards being a sort of "control board" and The Gaming Den being the meat-grinder), unless you put a disclaimer indicating how each forum tends to behave regarding homebrew. That's the closest thing I can see to Metacritic as possible (and to Rotten Tomatoes, as well).
The Wiki definitely needs some serious cross-promotion, however, for people to check it out. That should help, at least, to promote the Wiki beyond the alternative (which I'll refuse to mention on grounds of being the complete opposite of this Wiki). - T.G. Oskar 09:28, 18 June 2012 (UTC)
Tarkis, is there any particular reason for displaying the rating average in terms of categories, rather than just giving a number? If it's on a familiar scale (likely percentile if we stick with a binary system), then people will certainly understand it and be able to draw more information from it. Another thing I'd like to note about your proposal is that it might be better to calculate all averages faithfully, and have a "minimum votes required" criterion attached to Community Favorite status. That would distort the data less, and have the same end result.
But I'm going to have to continue agreeing with GW on greater granularity, and insist along with him that it is the better basis for a rating system. If we have a binary system, all we have is an indication of how likely someone is to like an article at all. We provide no information whatsoever about how much people like it -- which I think is far more valuable information, and which I think more directly correlates to overall quality. Toy Story 2 has 100% on Rotten Tomatoes. But few people would name it among the greatest movies ever made. It's just broadly agreed that it's quite good. (And Oskar, that is what I meant by comparing Metacritic and Rotten Tomatoes. Rotten Tomatoes works off a binary system of "do you like it or do you not", whereas Metacritic works off a normalized average of "how much do you like it". I don't think drawing a consensus from a forum would be feasible, and I don't think giving them any sort of implicitly elevated prestige as critics would be constructive.)
As for cutting out this conversation "full fucking stop," I agree with TF. Encouraging votes under a system likely to change is a waste of time and effort. We need to fix either this first and that second, or both of them simultaneously. Starting a new discussion heading for that issue might be productive. --DanielDraco 17:46, 18 June 2012 (UTC)
Any averages we do without 1) a large pool of raters or 2) an extremely granular ratings system are going to be very swingy. An article with 1 like / 1 dislike has a percent average of 50% and an article with 2 likes / 1 dislike has a percent average of 66%. An article with 1 like / 1 neither would have an average of 75%, while one with 2 likes has a completely boring 100%. These are not uncommon occurrences in rating, and the jumps between them suggest we don't need the granularity you want because we're not using it. Further, all of those numbers look really of-putting to someone who is used to percentage ratings. I don't want to help people pre-judge an article based on information that is too granular for its inputs. Even moving to a 5 point scale instead of the current 3 point doesn't resolve that, it just makes the gaps between jumps slightly smaller. It can work for metacritic because they have a more granular set of inputs and a greater number of them, but it's a bad fit here. Using granularity on the order of our ratings granularity doesn't have the same problems because we control the break points. The categories serve the same role as the percentages at metacritic, it's a thing you can look at to get a sense of the overall rating without being misled. Yes it's less information, but the information you want to give out is misleading or destructive so there's no reason to do it.
Also metacritic is a terrible example for what you want. It leans high and uses inputs that we don't. If anything, Amazon or Netflix with their 1-5 star ratings would be what you want to focus on, because that's what you're actually proposing. How useful do you feel their averages are when they have 1-2 ratings
As to calculating averages faithfully, not doing that was a conscious decision to deal with very small data sets. 2 perfect scores should not be reflected on the page in the same way as 4 or more perfect scores, not if we're trying to give people something simple to look at to understand how well the article is regarded. We could add a "based on X ratings" to the bottom of each score or just list the number of ratings or each type a page has received, but I see that as less helpful and requiring more work on the users part than just obscuring small data sets from the start and using categories. - Tarkisflux Talk 19:58, 18 June 2012 (UTC)
I really don't see the link between the swings present in a binary system and anything in a more granular system. Binary systems are far more swingy, because every vote is always at one extreme or the other. The jumps that happen with each vote suggest nothing except the fact that the span separating any two dissenting votes cannot possibly be greater. If one voter disagrees with another voter, the difference between them is always 100%. That will cause big swings from slight disagreements. More granular systems allow for less distance. If two users vote and disagree with each other, the odds are that the span between their opinions will be less than 100% of the entire range of options -- so it will, in most cases, jump less. The only way a more granular system can be as swingy as a binary system is if every user votes either the minimum or the maximum, with nobody in between -- a self-imposed binary. But they won't do that. There will be 2/5s and 4/5s. It will sway more gently, and zero in on a central point far more quickly.
And TF, frankly, I have to say that I don't tihnk "all of those numbers look really of-putting to someone who is used to percentage ratings" is a valid argument for what you're proposing. The numbers look bad, so we lie about them and say that they're other numbers? That's not serving the user at all. If people don't understand the notion of a sample size, a better accomodation would be to just not display any rating at all until a certain threshold is reached. That way we're not misrepresenting the data, but we're still avoiding the pitfall of less math-minded users thinking that one person's rating of 0/5 is the gospel truth. I do understand wanting to avoid publishing insufficient data. But the way to do that is not publish insufficient data. Distorting the data in any way does nothing but make it even less relevant. If it's not worth showing, don't show it. In fact, I would go so far as to propose this, as an element to whatever system we use with whatever level of granularity:
  • An article will not display voting data or an abstraction thereof until it has received 4 votes. Instead, it will display a note to the effect that more votes are needed. (The exact nature of this note and the prominence of it will be decided in the discussion below.)
  • An article is ineligible for rating-based categories such as "Community Favorite" until it has received 4 votes.
For the movie site analogy, I guess Metacritic isn't a great example. I just chose it as a counterpoint to Rotten Tomatoes, because they ultimately use the same data in different ways to achieve different results, and Rotten Tomatoes' way of handling the data is much like our current system. The pitfalls of RT are really the greater part of that analogy, anyway. But sure, let's use Netflix as the other side of the coin if that comparison is used again. --DanielDraco 18:11, 19 June 2012 (UTC)
The argument was not against a more granular system, which I don't actually care about. It was against a displayed rating average that was more granular than it's inputs. Moving to a 1-5 rating makes it better, as I noted, but it does not resolve the fact that a straight percentage system is full of data points that we are unlikely to ever use. That those percentages also come full of preconceptions and biases is just more reason to not use them. I'm less arguing against a 1-5, rounded to nearest .1 or .2, shown total than I am straight percentages, because that is a granularity closer to the scale we're actually using. But in no way is a derived shown value like I'm proposing misrepresenting data. I'm not giving as much as you want, because as much as you want is both useless and stupid given voting realities here, but I'm not saying untrue things with it or planning on hiding the way it's generated.
As to the rest, that's crap. You're actually proposing that instead of just dealing with small numbers of ratings in such a way that they don't over or under sell an article, we just don't display them at all. No rating at all on an article with less than 4? That's most things on here. That straight up doesn't work without major changes in rating behavior, and doesn't even look like an honest attempt to deal with the realities of rating on this site.
And as nothing new is being added at this point but bickering, I'm going to end it here and just write up a set of voting options. Granularity: More/Same. Rating:Category based on average, Average, or Numbers of each. If average: Weighted (minimum divisor), Minimum Required, or Straight. We have some things to decide on. - Tarkisflux Talk 20:13, 19 June 2012 (UTC)
Edit conflicted with Tarkis with this. Swinginess of a system average doesn't mean shit if we aren't displaying an average it in the first place, which is what I thought the case was with like/dislike, where the average is used only to calculate community favorite and not actually displayed in the author template (that display is likes/dislikes only). Surgo 20:15, 19 June 2012 (UTC)

No One Using Rating

Just bringing this up under a separate heading so it doesn't get lost among the noise. Nobody rates because there's no easy one-click solution to do so. I don't know if I'll have the time to program one. Maybe someone can find a canned solution on the mediawiki extension listings or something. If so I'll check it out. Basically such a solution should let the user click like/dislike, say why, then as its output drop exactly that onto the Talk page just as is done now. No difference in what the ratings are, just how they're inputted. Surgo 20:31, 18 June 2012 (UTC)

It would also need to update the author template of the article page, or set a property on the article page that could be referenced. Otherwise we don't get the visibility that we want. - Tarkisflux Talk 20:38, 18 June 2012 (UTC)
How difficult would it be to program this kind of functionality? I'm done with finals now, and thus, I could probably do it if someone told me what I'd need to learn in order to make it happen. - MisterSinister 20:38, 18 June 2012 (UTC)
There are some extensions that handle the numerical voting bits, some complete and some experimental (see this useful category page), but only 1 comes with an attached comments box (this one). None of them really work for what we want though. So it looks like we'd need to code from scratch or sacrifice some desired functionality. - Tarkisflux Talk 20:51, 18 June 2012 (UTC)
Actually, it doesn't need to edit the author template specifically. We could do a new template for the page called {{Raters}} or whatever that set properties, and the author template could just look them up in the page's metadata. Or if there's a way to do a direct sql query from a template (which I think there is, though I'm not sure how), we could skip editing the author page entirely and have the ratings code just drop the lists of raters in a table somewhere and query that. We just need to be able to update such a table when ratings are changed / removed. - Tarkisflux Talk 21:16, 18 June 2012 (UTC)
Difficulty depends on how familiar you are with PHP. You'd need to learn the MediaWiki codebase. I highly doubt you can make a SQL call from a simple template (it would be crazy easy to DDOS the site let alone all the security holes you could imagine), but you can certainly do it from a template that calls a function where the function itself is implemented in PHP code. EG {{#if: }}, which doesn't use SQL but you get the idea. Surgo 21:45, 18 June 2012 (UTC)
Since I wrote that I remembered you can query the properties of other pages and list them out (and then do string tricks on them). So if we turned on smw in the talk space, we can set ratings with the rating template. And then query those values to do stuff on the article page. So if we wanted to simplify things without worrying about writing a form equivalent to append things to a page, we can reduce it to a single template addition on talk (with some backend changes). - Tarkisflux Talk 22:00, 18 June 2012 (UTC)
Ok, so SMW in talk is turned on now and the author template has been updated to make use of it. All that is needed to rate something is to put the template up on the article's talk page. We probably don't need to worry about any extensions at this point, though something like a preload that allowed you to append a pre-formatted block of text to an existing section in a page would be nice (since the preload can only create new pages or new sections at present). - Tarkisflux Talk 18:44, 19 June 2012 (UTC)
If we want to actually get a volume of ratings, we absolutely do need a one-click system for doing so. Surgo 20:11, 19 June 2012 (UTC)

Voting

Another wiki vote. Yay.

Rules:

  • Add your signature to vote.
  • We're doing instant run-off in 2 categories: Granularity and Rating Display. You get 1 vote for your first choice in each category, and 1 vote for your second choice in each category.
  • Keep comments out of the voting lines for legibility.
  • Voting runs for at least 2 weeks.

Granularity

Current Like/Neither/Dislike

First Choice

Second Choice

1 to 5

First Choice

Second Choice

0 to 10

First Choice

Second Choice

Rating Display

Derived Category

Ex: Most users liked this article / No clear feelings are obvious for this article / Most users disliked this article Note: If derived wins out, there will be a followup discussion to determine the categories and placement requirements. Holding off for now, because it's more voting that might not even matter.

First Choice

Second Choice

Average

Ex: 75%, 3.2 Note: If average wins out, there will be a followup vote to determine the version displayed and the method of calculation. Holding off for now, because it's more voting that might not even matter.

First Choice

Second Choice

Straight Numbers

Ex: 2 Likes, 1 Dislike; 1 user rated this a 4, 1 user rated this a 5

First Choice

Second Choice

Comments and whatever

Categorization is my preferred method, but if it or Average wins I would want to display the straight numbers in a spoilered box at the bottom of the ratings display, to make it easier to see how the numbers fall out at a glance. - Tarkisflux Talk 20:52, 19 June 2012 (UTC)

When we only get 1 to 5 votes, we could even potentially show who likes and dislikes the page. --Foxwarrior 21:32, 19 June 2012 (UTC)

Just reiterating: we tried scales on multiple occasions, at no point did it ever work. I don't see it working now either. 108.6.239.14 23:33, 19 June 2012 (UTC)

Retrieved from "https://dnd-wiki.org/w/index.php?title=Dungeons_and_Dragons_Wiki_talk:Rating_Articles&oldid=159619"