What are the best practices for tagging on Gaming.SE?

Question

I think we've kind of been dancing around this question, and I think that it's a question that really needs a definitive, solid answer that everyone agrees on. I can't find a meta post (perhaps one already exists and should be updated? Please correct me!) that really has this locked down. As pretty much a noob here, perhaps it's my place to ask these really noob-ish questions.

What problem(s) does tagging solve on Gaming.SE?

I think this is sort of the first question we should be answering. What is a tag for? What is a tag explicitly NOT for? What is the goal in picking tags on this site? Organization? Filtering? Subscriptions? Some combination?

How should questions be tagged on Gaming.SE?

How should we use the tag system properly to meet the problem we're trying to solve? "Power Users" on Gaming.SE seem to do a lot of tag maintenance. What suggestions should we give users & examples should we set to minimize the need to do this maintenance?

For reference, I found the following posts, which don't really seem to reflect recent guidelines:

Why does Gaming.SE use tags differently from other SE sites?

Tagging Suggestions

What should be our battle plan for tagging strategy and tactics?

Finally, I apologize if I'm overstepping or if I'm doing this wrong. That's the other nice thing about being a noob, you get to make silly mistakes now and again ;) I hope you can forgive me if this is one of them.

Answer 1

The first things to consider are the two core points of tags. These are the core essences that must be respected above all else, and we cannot ignore these.

The first core point of a tag is it must describe the subject matter of the question. In other words, if you look at a tag, it should be telling you what the question is about, not what the question is or contains. For example, our primary tag case is game title tags - these tell you what video game the question is asking about. Tags which tell you what the question is or contains are known as meta-tags - they tell you information about the question, but not what the question is asking.

Secondly, the tag helps users filter questions by their subject matter. We have the ability to search through tags, and to mark tags as favorite/ignored. These are all facilities to help guide our users to the questions they want to see. Without tags, users are just swimming in a bed of questions without guidance.

---

Non-gameplay questions, such as those about console hardware and such, are pretty easy to tag. I don't think we've really run into any major problems with those. Our biggest problems come from tagging questions about gameplay.

In the very vast majority of gameplay questions and the like, a single tag, the game title, is enough. Yes, you could identify that "this question is about solving a particular puzzle", while "that question is about how a certain ability works in the game" by tags, but in general this level of granularity is simply far too strong. This is because in most cases, no one cares about that level of granularity. No one sifts through it, no one regularly groups these questions together, no one chooses to specialize or avoid that wide of categorization. Remember, a core point is that a tag must help filter, so a tag whose filter is effectively meaningless (often due to being too broad) is a tag that must be avoided.

Extra tags which filter beyond the game tag, what are often referred to as sub tags, are only useful if they represent a meaningful subdivision of the game's question material. As tags help prevent our entire question base from being a directionless swamp, subtags help prevent large tags from being equally overwhelming. They exist where people have strong reason to either pay heed or overlook certain questions. And they also exist where this subdivision can be identified by a keyword. The best examples of these come from our most accepted sub tags: the Starcraft race tags terran zerg protoss and Minecraft's own tag minecraft-redstone.

The intense strategy of Starcraft and its successor make the identification of race a veritable keystone in the development of successful gameplay. The tags allow players to find questions of their interest - they effectively give the software the ability to know that a question about Mutalisks is about Zerg, that Stalkers belong to Protoss, that Siege Tanks are built by Terrans. Meanwhile, redstone may be just another part of Minecraft, but its depth and distinction is so great that it really represents a different class of gameplay altogether. Both of these are examples where the tag truly helps - it helps identify a distinct body of knowledge that is useful to have divided from the rest of the questions in that game. And furthermore, it helps group these questions in a natural and meaningful way that cannot always be grasped by text search alone.

Answer 2

Let me see if I can sum up what everyone's been saying. I think this is probably going to take a few revisions to get right, but hopefully I'm at least getting closer. I'm going to make this sound authoritative, but I'm just summing up the consensus, so if the consensus is different let's make changes.

The Problem

There are too many questions on G.SE for users to meaningfully parse. Keyword search relies on machine intelligence to sort the questions, and doesn't solve all of the sorting and filtering problems users encounter. Users should have some way of filtering the questions to just those they're interested in. This filtering should be possible in any SE use case (ie, whether browsing, feed reading, searching, or in pushed notifications) This system should be easy to use (for both normal users and people doing maintenance) and yet powerful enough to work in a variety of situations.

The Accepted Workaround

Tags are a way for users of the site to provide intelligent meta-data to a question. They allow users to categorize a question in broad terms. They form 'buckets' into which we can place questions. The trick to tagging is to creating buckets that are both large and small enough to be useful. A tag with too many questions begins to experience the same problem as a Q&A system without tags. However, a tag with too few questions becomes to specific and esoteric to be useful.

To avoid these issues, our tagging policy is as follows:

• Tag the platform if there's no game specified, or the question is platform-specific.
• Tag the game name. If a question is game-specific, use that as a tag. Sometimes game names are long, and that's OK, we can work together to figure out a shorter name that makes sense.

• Sub-tag the game if the game name is insufficient - In situations where the game name creates too broad a tag, we may wish to create tags that start with the game name as a prefix, and then categorize other tags based on this prefix. For instance, if there are 1,000 starcraft-2 questions, we might have a zerg tag, among others. Avoid duplicating information in subtags when they are already unambiguous.

• Be cautious about non-specific tags. - non-specific tags transcend games and can be tricky. A non-specific tag is something like "weapon". These tags tend to only make sense with more context. There's some debate as to whether or not tags that can't stand alone on a question make sense.

• Avoid meta tags. - "spoiler" is an example of a meta tag. It describes the content of the question, rather than the content.
• Avoid over-tagging. - When in doubt, bucket the question broadly. (This one is somewhat debatable, see below)

Moderators and power users will work with specific situations in order to address problems as they arise. Use this as a guideline, but understand that this is largely an ad-hoc process and it is still evolving. Use your best judgement, and we'll adapt as we need to. Recognize that this is subjective, and ad-hoc, and things will change and edits will be applied. Deal with it!

---

Now to go off on a different tangent...

Drawbacks and Observations

If I could have 10 minutes to sit down with the designers/implementers of this system, I would bring up the following points:

• Sub tagging, and game name tagging in the manner above seems contrary to the design of tags. Tags are designed to be short (25 characters does not appear to be a technical limitation), and the number of tags is intentionally small. We're abusing your system here.
• Sub tags create hierarchy in a system that doesn't support it. We're creating a "game-weapons" tag to sort through too many "game" tagged questions. This is not what the system was designed to do.
• Sub tagging creates a bunch of work for question maintainers. Sub-tags are not intuitive, so applying them is ad-hoc at best. When to apply sub-tags is also problematic. We either overtag at the start and then merge/delete later, or we go back over a large swath of questions after the fact and re-tag. Both are messy and time consuming.
• Searching seems to have issues. Some have pointed out that plural words don't match singular words, for instance. Some of the things we're doing in tags is to work around search issues.

I'd like to point out that these aren't problems specific to G.SE, although it has been argued that they're more acute here. If I had time to ask just one question, it would be: Tagging alone, in its current form, is insufficient (or unclear) to solve the problem statement above. How do you propose we adapt our behavior (or adapt your system) to solve this problem?

I could propose longer-term solutions to this problem, but I don't want to get mired down in a technical debate. On top of that, I'm not the designer, architect, master implementer, or even a guy who has spent 5 minutes with the code. I'd love it if you gave a crap about my opinion on the subject, but I'm not going to presume that you do. As a user, I think the best I can do is file a clear bug report :)

Answer 3

The best practices are: there are no best practices. Tagging happens. Deal with it.

Tagging is performed mainly by users with no experience with the site, and never was meant to be perfect. You should then not worry about attempting to perfect it.

We can steer this process into our will in two things:

• Editing. This does not scale. We are never going to retag all questions on the site. Even if we do, more questions will come, ideas on how we should do tagging will change, the software will change, and we'll be back to square one.
• Question tag counts. Users will tentatively try to write words in the tag area. If we have a tag matching that word, they'll use it. The more questions a tag has, the more prominent its match suggestion will be. We want to keep bad tags to a low count and good tags to a high count, so that good tag suggestions come first.
• Nuking bad tags. Users can't create tags. Okay, 99% of them can't. If we have a bad tag with lots of questions, have it nuked. Blacklist it if possible. (Moderators can't do this.)
• Merging. A simple way to boost a tag's question count is to merge it with other, similar tags that will pop up naturally from the 1% that can make tags.

There's a problem here though - some tags are better than others. In Stack Overflow, the "top" level of tagging would be a language tag. Here, they're game tags. It would be important to make those prominent where space is at a premium - title pages, twitter, etc., but we can't have this right now. What we do have is the tag with the most questions being featured. This means more generic tags will move ahead.

This isn't, countrary to popular belief, a problem only Gaming has — to find an example on Stack Overflow, just take any common problem category and any minor language. Example: the go tag (×536) has no change over the [strings] tag (×17,986), so when using both people have resorted to just write "in go" in the title, or even more crudely prefix the titles with "golang -". Where people haven't done this, we get a terrible question title (and tweet).

Can you answer this? bytes.Split separator as []byte(“…”) http://bit.ly/whatever #string

It's pretty much a lose-lose situation.

Is the solution slashing and hacking tags? Certainly, if this question was tagged go go-string then go would be the top tag. And certainly, strings in go and strings in c are very different animals (if they aren't, just consider strings in Python or Java).

Yet, no matter, you can still ask questions about strings - for example, about encoding or decoding. Besides, strings in C aren't different from (raw) strings in C++ or most other languages. There are only so many kind of strings out there - zero-terminated strings, length-prefixed strings, etc.

Contrast with Gaming. Every game out there has its own leveling system. Can you ask questions that are about generic leveling systems? Certainly - but they'd be design questions that are off-topic here.

One of the sanity checks SE does before migrating a question from a site to another is checking if any of the tags on a question are existent on the target site. If this check fails, the question is simply closed off-topic without being migrated. (I'm talking about close votes - moderator initiated migration overrides this.)

After all, why should we have tags for topics that are outside our scope? Discussing weapons in general is a gamedev topic, and thus is off topic here. If you want to discuss them in the context of a game, however, that starts to make sense and becomes our own field.

_{Does that mean that we should start covering design questions? Probably; I'd love that personally, although we lack the experts. And that's a whole different can of worms.}

Even if we do (and we do, and there's very little we can do about), as noted above, those tags need to be kept under control, before they start covering tags about topics we do cover. And so many of the topics we do cover are at 0, 1 or 2 questions each; this is probably what makes Jeff so scared about accepting space-bunnies-must-die as a valid tag. Is it a game? Yes. Do we cover that topic? Then we have a tag for it.

Do we cover weapons in general? No. If somebody were to ask "WHAT GAME HAS THE GREATEST FIREBALL SPELL EVER?" (A: Trine), would he find tags to use? Yes: weapons. He'd tag the question anyway by the power of the dancing bunny, but the fact that we have a tag that is just about weapons no matter of the game we're talking about does then trigger legitimate confusion: "what do you mean off-topic? I can't ask a question about weapons? You have a tag for it, and a hundred friggin' questions!".

"Oh, but badp, what about boss-fights, or achievements! Do we really need a subtag for each and every single game?" No, we don't. We've already established that those tags wouldn't be protected against pruning under the proposed "important tag" system. If a game doesn't have more than one (or two, or five) questions about bosses, subtagging them as such is kind of useless.

"But you yourself argued that boss-fights is a good tag because you can ask and answer questions about playing the generic boss fight! That's not design!" I don't honestly know if talking about weak points of common boss fight design is itself design or not and thus on- or off-topic here (I'd definitely go for on-topic, especially if we could get an answer as epic as that one). Again, though, such a question is the needle in the hay, and if taken out of context ("HOW DO I BEAT BOSS") meaningless. The context in which you put this question will give you a better tag to use (or create) than boss-fights.

This is why I thus believe that, at least for large enough games, we should go towards prefixes in non-game tags for games we feel will reach a large enough size to warrant them. If a game doesn't turn out to be as huge as we thought, and the prefix we've chosen (for space reasons) gets in the way of newer prefixes, we can merge them out of existence with their game tag cleanly and swiftly. This also means we get to use abbreviations in context-ed subtags, which helps with the tag lengths. I can hardly imagine going longer than skyrim-boss-fights (if, again, there are even enough to warrant this tag in the first place).

One problem remains: the homepage. "minecraft minecraft redstone" or "team-fortress-2 tf2 spy" is an unsightly, spammy mess. I hope that by progressively abandoning contextless tags, though, we'll still get to address the problem enough to make this measure unnecessary.