Class Balance in Tabletop Role-Playing Games: Kinds of Fun

This is the first in a series of posts about class balance in tabletop RPGs. This post describes the “kinds of fun” model of class balance.

Sean K. Reynolds says that you can’t balance classes in a tabletop RPG except on the basis of damage. Many players and GMs agree (in my experience, anyway).

They’re wrong.

There are two ways of thinking about class balance in tabletop role-playing games (a.k.a. “pen-and-paper RPGs”) that solve, or cut through, the problems SKR describes, and that describe player views and behaviors fairly well. This post talks about one of these ways of thinking: the “kinds of fun” model. (A subsequent post will describe a lower-level model of class balance, the “evaluative metrics” model.)

I. What Is This… “Fun”… Thou Speakest Of?

What’s this about “kinds of fun”…?

Well, for instance, here’s a paper that describes eight types of fun that people get from games, and here’s a blog post discusses the paper and its claims.

It’s not necessary, though, that we buy into the paper’s specific claims. We don’t need to use that exact taxonomy of fun, either. All we have to agree on is the intuitive idea that people can derive different sorts of fun from the same game. This idea is pervasive in discussion of tabletop gaming. “Role-playing” vs. “roll-playing”; “intrigue”, “drama”, “exploration”, “puzzle-solving”, and “combat”; “Timmy”, “Johnny”, and “Spike”; many taxonomies, categories, and classification schemes have been invented, but the gist is always the same: a single game can provide more than one kind of fun. Different people can want, and have, different kinds of fun; or the same person can want, and have, different kinds of fun at different times, or even at the same time. All we need is the idea that there’s different kinds of fun to be had from the game we’re looking at.

Note: I’ll be talking about, and using examples from, Dungeons & Dragons and similar games, but the ideas in this post apply much more broadly; supplying the details for other RPGs is an exercise for the reader.

II. Fun For the Whole Class

The other idea we’ll need is that there’s some meaningful mapping from character classes to kinds of fun. That is, we need the idea that a character class can provide one or more kinds of fun, or, in other words: the idea that for at least some of the kinds of fun that it’s possible to have while playing this game, you need for your chosen character class to have certain properties. (There may be kinds of fun that the game can provide, that have nothing whatever, even in principle, to do with your choice of class—for example, the fun of socializing with your friends. We exclude such kinds of fun from our discussion; we’ll focus only on those types of fun that have at least some dependence on what your character’s like—and more specifically, to what class your character belongs.)

Again, we don’t need to commit to any specific taxonomy of fun, and we don’t need to specify the mapping from character classes to kinds of fun. It doesn’t need to be a one-to-one mapping, where each character class can provide one, and only one, kind of fun; nor does each class need to have a monopoly on the kinds of fun it can provide. The mapping can be many-to-many; and we certainly don’t exclude the possibility that, in some particular game, each class has all of the properties necessary to provide all of the available kinds of fun. We also don’t need to say anything about just what these necessary properties of character classes actually are—particular game mechanics, no doubt? or maybe it’s something to do with “flavor” or “fluff”?—we can be almost entirely agnostic about this. Say only that there are such properties—that is enough.

We’re now armed with the two ideas we need: that there are different kinds of fun to be had with this game, and that at least some of these kinds of fun have some dependence on the properties of the character class you select.

III. Hit Things Real Hard

Not even Sean K Reynolds denies that classes in D&D can be, and often are, compared on the basis of damage output in combat, and that players seem to care about such comparisons. Can we fit this uncontroversial truth into the framework we’ve set up so far? (If we can’t, we’re in trouble—it only gets trickier from here!)

Sure. We can note that there’s a kind of fun that is provided by playing a character who does a lot of damage in combat. (What is the nature of this sort of fun, exactly? What are its psychological properties, etc.? We don’t need to know any of that! We only need to know that this sort of fun exists—no problem at all, since it’s easy to find many examples of players who specifically enjoy their characters doing lots of damage in combat.)

We can then look at each character class, and ask: does this class let me do lots of damage in combat? If yes: then this class provides the doing-lots-of-damage type of fun. If not, not.

But wait! We’ve actually run into a complication. Can you spot it, in the above two paragraphs?

It’s the phrase “a lot”—as in, “a lot of damage”. What the heck is “a lot”? 2? 12? 200? Per round…? Per attack…? Enough to kill a dragon in one hit? Ten hits? Well, ok, if you can kill a dragon in one hit, I think that most people would agree that you’re doing “a lot of damage”. But in practice, “a lot of damage” most often means “a lot of damage, compared to other characters”. That is: however much damage other characters can do—whether it’s 2 or 12 or 200 points, per round or per attack or whatever—if you do “a lot of damage” compared to that, then you’ll probably say that you do “a lot of damage”, period.

(Note that, in general—outside the context of this analysis—what I’ve just described is not the only possible way to define “doing a lot of damage”. For example, you can define it relative to encounter design: you can say that a character does “a lot of damage” in combat if he does so much damage that he enables his party to defeat combat encounters more quickly than the game designers intended or expected them to be defeated. But the definition given above—i.e., defining “a lot of damage” relative to other players characters—is (I claim!) the correct definition when discussing whether or not a player is experiencing the sort of fun that is associated with doing a lot of damage in combat. If you doubt this, go to any major RPG-related discussion forum, and count how many players complain about not doing as much damage as other PCs; and compare that to how many players complain about not doing enough damage to defeat encounters as quickly as intended.)

In other words, “how much damage does my character do—a lot, or not a lot?” is not a one-place function; rather, it’s a two-place function (at least!). It’s defined relative to—well, to several other factors, actually, but first and foremost, relative to other player characters.

Note, again, how many of the details we don’t have to specify—and thus, how broadly we can define, and apply, our analysis. What is “a lot, compared to other characters”, anyway? Twice as much as the next most damaging character? 50% more than average? At least as much as anyone else? We can be agnostic about this, too. All we need to know is that this idea of “a lot, compared to others” is coherent and applicable in practice. That’s hard to deny. I think that nearly all players will agree that “a lot of damage, compared to other characters” is a pretty intuitive idea (even if they disagree on the details!).

The takeaway from this example is: when we talk of “properties of a character class” (that determine, in some way, in whole or in part, what kinds of fun may be had while playing a character of that class), at least some of those properties are defined, in part, relative to other player characters.

IV. The Fighter and the Shoemaker

We’re almost ready to build our class comparison framework. First, we need a couple of toy examples.

For these examples, we’re going to assume that every player selects a different character class. (In real games, multiple players sometimes play characters of the same class. Then again, in real games, there are all sorts of complications: multiclassing, prestige classes, feats, etc. The model I’m describing is a simplified one—but not nearly so simplified as to be useless. If you think some particular property of real games invalidates what I’m saying, tell me so in the comments.)

Imagine a game system, much like Dungeons & Dragons, but with only two classes: the fighter, and the shoemaker.

The fighter is very strong and tough, can use any kind of weapon, and knows all sorts of fighting techniques.

The shoemaker makes shoes. He can’t fight.

(How useful is the ability to make shoes? What else can the shoemaker do—or the fighter, for that matter? We don’t care about any of that, for now.)

A game played using this system will have two characters: one fighter, and one shoemaker. Now, we ask: does the fighter do a lot of damage?

Yeah, he does. He certainly does a lot more than the shoemaker (what’s that guy going to do—throw a shoe at the enemy? We’re fighting dragons, not U.S. presidents), and that’s what matters. And so, in this game system, the fighter class provides the sort of fun that’s based on doing lots of damage in combat.

(Does the fighter class provide any other kinds of fun? Does the shoemaker class provide any kinds of fun at all? These questions, too, we set aside for now.)

V. World of Paladincraft

Now imagine a slightly modified version of this very same game system. Again we’ve got our fighter class and our shoemaker class, just as before, but now we’ve also got a third class: the paladin. The paladin is actually quite similar to the fighter: he’s very strong and tough, can use any kind of weapon, and knows all sorts of fighting techniques. But unlike the fighter, the paladin can call upon the holy power of the cosmic forces of Good. This holy power has many benefits, among which is that every single one of the paladin’s attacks do triple damage. A paladin, in other words, will always be doing three times as much damage as a fighter (of the same level, etc.).

A game played using this system can have two or three characters: a fighter and a shoemaker, a paladin and a shoemaker, a fighter and a paladin, or one of each of the three classes.

We ask again: does the fighter do a lot of damage?

If no one’s playing a paladin—if a game contains just a fighter and a shoemaker—then the situation’s exactly the same as before. The fighter does a lot of damage. His player is having the “doing a lot of damage” sort of fun.

But what happens when we do have a paladin?

(Remember, now, that we’re not asking whether the fighter’s doing enough damage to defeat encounters; or how many hits it takes him to kill a dragon; or any such thing. What we’re after, here, is whether the fighter’s player is experiencing the sort of fun that comes from “doing a lot of damage in combat”—whatever that means. And as we’ve noted, what it means is: “does the fighter do a lot of damage, relative to the other player characters”.)

I think it’s clear enough that, in any game in this system where there’s a both a fighter and a paladin, the fighter can’t be described as doing “a lot of damage”. I mean—the paladin’s doing three times as much! All the time! The fighter’s player will not be having the sort of fun that comes from doing a lot of damage in combat. And the blame for that Fun Deficit may be placed squarely upon the shoulders of the fighter class. It’s exactly the properties of the fighter class—specifically, what class features it has and which it lacks, compared to the paladin—that are the sole reason why the fighter’s player isn’t having this particular sort of fun.

And now we’re ready to generalize.

VI. Fun in Three Dimensions

In this model of class balance, there are three variables that we care about. We’ve got a, which may be any one of the character classes in the game system we’re examining. We’ve got b—any other class in the game system. And we’ve got k—a kind of fun (and specifically, a kind of fun that, in this game system, has some dependence on the properties of your chosen character class).

For every possible combination (a, b, k) of two different classes and a type of fun, we ask: given that a character of class b is present in the game, will a character of class a experience fun of type k? (If the answer is “yes”, we say that class a is capable of providing fun of type k in the presence of class b.)

For example, in the hypothetical game system described in the previous section, if

a = fighter
b = paladin
k = the type of fun associated with doing lots of damage in combat

then the answer is “no”: the fighter class does not provide this sort of fun in the presence of a paladin.

By enumerating either b or k, we can look at two different aspects of class balance. (In the next section, we’ll put it all together into a unitary model.)

First, we can hold a and k constant while enumerating b. That is: given a class a, and a kind of fun k, in the presence of which of the other classes can and can’t class a provide fun of type k?

For example (sticking to the same “fighter, shoemaker, paladin” game system we’ve been looking at) we can ask:

In the presence of which of the other two classes can the fighter provide the type of fun associated with doing lots of damage in combat?

Here, a is the fighter class; k is the type of fun associated with doing lots of damage in combat; and the possible values of b are the set: { shoemaker, paladin }. We can arrange the answers for each of the values on a table:

Fun based on doing a lot of damage
	Other class (b) in the game
Class (a)	shoemaker	paladin
fighter	yes	no

The fighter can provide the type of fun associated with doing lots of damage in combat in the presence of the shoemaker, but cannot do so in the presence of the paladin.

Second, we can hold a and b constant while enumerating k. That is: given two classes, a and b, what kinds of fun can class a provide, in the presence of class b?

Still sticking with our “fighter, shoemaker, paladin” game system, we now need to enumerate all the kinds of fun that a) this game system can provide, and b) depend, in some way, on the properties of a player’s chosen character class. We’ve got one kind of fun already: the sort of fun that comes from playing a character who does lots of damage in combat. Let’s now postulate another kind of fun: the fun of playing a character who can run a business. Again, there’s no need at all to delve into the psychological nitty-gritty of this sort of fun; we’ll just take it for granted that this is a real thing that some players are going to enjoy, and that this game system’s mechanics and structures can, in principle, support. Let’s also suppose that there’s no generic skills system in this game, nor any other non-class-dependent way of engaging in commerce; in this game, either your class provides your character with the abilities you need to be a businessman, or you’re SOL. (This, of course, is exactly the case with combat, so our assumptions here are quite symmetric.)

Who can run a business? The shoemaker can; he makes shoes and sells them. Meanwhile, neither the fighter nor the paladin have any business savvy whatsoever. (They can try to start a fighting-dudes-for-hire operation, but they’re such dunces that they’ll get swindled immediately and will, if anything, actually lose money; or they’ll run afoul of the kingdom’s trade guilds and licensing laws; etc.)

So, for example, let’s look at the fighter (that’s our a) and the shoemaker (b). The possible values of k are the set: { the kind of fun having to do with doing lots of damage in combat, the kind of fun having to do with running a business }.

Fun based on doing a lot of damage
	Other class (b) in the game
Class (a)	shoemaker
fighter	yes

Fun based on running a business
	Other class (b) in the game
Class (a)	shoemaker
fighter	no

In the presence of the shoemaker, the fighter can provide the sort of fun that’s based on doing lots of damage in combat, but cannot provide the sort of fun that’s based on running a business.

VII. Balance of Fun

At this point in the post, it might seem like we’ve lost track of the initial goal. Weren’t we going to talk about class balance?

But look what we’ve just done, in the second half of the previous section. We’ve taken two classes, and we’ve asked what kinds of fun the first class can provide, in the presence of the second class.

Let’s now define the following relation:

Class A is strictly worse than class B if, and only if, there are zero kinds of fun that class A can provide in the presence of class B.

This is a '’very conservative criterion. It calls one class (A) strictly worse than the other (B), only if the presence of a character of class B prevents a player of a character of class A from having any of the kinds of fun that have anything at all to do with the properties of a class. It’s hard to imagine anyone arguing with this definition—unless it’s to call it too'' conservative!

For example, in our hypothetical game system (where the set of classes is { fighter, shoemaker, paladin } and the set of kinds of fun is { the kind of fun that comes from doing a lot of damage in combat, the kind of fun that comes from running a business } ), the fighter is strictly worse than the paladin. In the presence of a paladin, the fighter class cannot provide any of the available kinds of fun.

I think we can safely say that, if one class is strictly worse than another class, then those two classes are not balanced. We don’t know what conditions might be sufficient for two classes to be balanced with each other, but this one is necessary: that neither class be strictly worse (by our very conservative definition!) than the other.

(Consider your favorite RPG system. Can you think of any two classes for which this criterion fails?)

VIII. The Fun Matrix

We are now prepared to unify everything we’ve established so far into a single model. This model will let us ask several sorts of questions about class balance in a game system:

1. For any given pair of classes (a, b), are either of these classes strictly worse than the other?

If the answer’s “no”, then these two classes are at least potentially balanced with each other; if the answer’s “yes”, then they’re definitely not balanced.

2. For any given class a, is there any class b such that a is strictly worse than b?

If the answer’s “yes”, this is a potential pitfall. The players, and certainly the GM, should be aware of this fact. (How to address this pitfall is, of course, a large and complex topic.)

3. For any given class a, is there any class b such that a is not strictly worse than (i.e., at least potentially balanced with) b?

If the answer’s “no”, then it’s hard to avoid the conclusion that class a is fundamentally flawed. A GM would do well to discourage players from selecting such a class.

4. For any given class a, what types of fun can this class provide unconditionally (that is, regardless of what other classes are present)?

Any player who cares about a kind of fun that class a can provide only conditionally (i.e., a kind of fun that class a can provide only if certain other classes are not present), should be apprised of this fact before choosing class a.

5. For any given class b, what constraints does the presence of class b impose on the ability of other character classes to provide the various available kinds of fun?

This is the question that’s at the heart of “party balance”. Much can be said about it, but that’ll have to be another post.

* * *

And now, the model. It’s quite simple—nothing more than a three-dimensional array (one dimension for each of the three variables defined in section VI). Two of the dimensions are the set of all character classes; these correspond to the variables a and b. The third dimension is the set of all the relevant kinds of fun (the variable k). Each cell of the array represents an (a, b, k) tuple. The array as a whole fully enumerates all possible such tuples.

Depicting a three-dimensional array is tricky, but you can slice it up into layers (each one cell thick), and lay them out as ordinary (two-dimensional) tables. Here’s the Fun Matrix for our toy example game system:

Fun based on doing a lot of damage
	Other class (b) in the game
Class (a)	fighter	shoemaker	paladin
fighter	—	yes	no
shoemaker	no	—	no
paladin	yes	yes	—

Fun based on running a business
	Other class (b) in the game
Class (a)	fighter	shoemaker	paladin
fighter	—	no	no
shoemaker	yes	—	yes
paladin	no	no	—

Each layer of the array is one of the kinds of fun. Rows and columns are classes. (Imagine the layers stacked one on top of the other.)

IX. Using the Fun Matrix

Armed with the Fun Matrix, let’s try answering those five questions we asked in section VIII.

1. For any given pair of classes (a, b), are either of these classes strictly worse than the other?

Recall the definition we gave: “Class A is strictly worse than class B if, and only if, there are zero kinds of fun that class A can provide in the presence of class B.”

To determine whether one class (A) is strictly worse than another (B), we look at every layer of the Fun Matrix, and in each layer, we look at the row for class A, look over to the column for class B, and see what it says. Let’s try it:

(fighter, paladin)—Can the fighter provide the first kind of fun (doing a lot of damage) in the presence of a paladin? No. Can the fighter provide the second kind of fun (running a business) in the presence of the paladin? No.

It seems that the fighter is strictly worse than the paladin. Unfortunate.

(It turns out that in our model, it’s possible for class A to be strictly worse than class B, and for B to be strictly worse than A. Is that a bug, or a feature? I think it’s a feature. If this happens, it’s likely a good indication that neither class is very good. They should probably both be redesigned.)

The paladin, on the other hand, can provide the first kind of fun in the presence of the fighter, so the paladin is not strictly worse than the fighter.

Similarly, we can see that neither the fighter nor the shoemaker are strictly worse than the other, and ditto with the shoemaker and the paladin.

2. For any given class a, is there any class b such that a is strictly worse than b?

As we’ve just seen, yes, this is the case for the fighter: it’s strictly worse than the paladin. (Neither the paladin nor the shoemaker are strictly worse than any other class.)

3. For any given class a, is there any class b such that a is not strictly worse than (i.e., at least potentially balanced with) b?

The fighter is potentially balanced with the shoemaker. The paladin is potentially balanced with both the fighter and the shoemaker. (Of course, we already know that the paladin isn’t actually balanced with the fighter; the fighter is strictly worse.) The shoemaker is potentially balanced with both the fighter and the paladin.

4. For any given class a, what types of fun can this class provide unconditionally (that is, regardless of what other classes are present)?

To figure this out, we look at each layer of the array; in each layer, we look at the row for class a; and we see if there’s nothing but “yes” entries in that row. If so, then class a can provide that kind of fun unconditionally. If there are any “no” entries in class a’s row in the layer for that kind of fun, then class a can provide that kind of fun only conditionally (i.e. only if certain other classes aren’t present).

So we see that the fighter is not capable of providing any of the kinds of fun unconditionally. Meanwhile, the paladin can provide the first kind of fun (doing a lot of damage) unconditionally, while the shoemaker can provide the second kind of fun (running a business) unconditionally.

5. For any given class b, what constraints does the presence of class b impose on the ability of other character classes to provide the various available kinds of fun?

Answering this one is a bit tricky. For each layer of the array, we’ve got to look at the column for class b; see if there are “no” entries in that column for any of the rows; and then, see if that row has any “yes” entries at all. (That is, what we’re looking at is whether class b makes any other class (a) unable to provide a certain kind of fun, which class a could potentially provide if class b weren’t present.)

We can see that the presence of the paladin prevents the fighter from providing the first kind of fun (doing a lot of damage). (Note that, even though the shoemaker can’t provide that sort of fun in the presence of the paladin, this is not at all a constraint imposed by the presence of the paladin, since the shoemaker cannot provide this kind of fun in any event.)

Meanwhile, neither the fighter nor the shoemaker impose any constraints on the ability of any class to provide either of the available kinds of fun.

X. Interpreting the Fun Matrix

Have we arrived at a definitive theory of class balance?

Not quite. But here are some things I think we can say about a game system, on the basis of our analysis so far. We’ll start with the obvious, and gradually move on to the more speculative.

If one class is strictly worse than every other class, that class is broken.

It’s hard to argue with this one. Remember: by our definition, one class is strictly worse than another only if the presence of a character of the second class prevents the first class from providing any of the available kinds of fun. That’s pretty bad. And we’re suggesting a scenario where a character class can’t provide any kind of fun, in the presence of any character of any other class! Terrible. This is a code red scenario. No game system should ever be ok with this.

If one class is strictly worse than another, this is a problem.

This isn’t necessarily problematic in all cases, since it may be possible (depending on the game system) to have games which don’t contain characters of both of the two classes. But even then, it’s a precarious situation. I think that if one class is strictly worse than another, one or both of the classes should be modified, or one should be removed entirely.

On the other hand, if there’s some class that a lot of the other classes are strictly worse than, that’s a problem too.

Suppose you’ve got a basically well-balanced game system, but then you add a new class; and in the presence of that class, a bunch of other classes suddenly can’t provide any kind of fun. Hmm. Something’s gone wrong, certainly (although it’s not necessarily obvious what—remember that we’re not talking about measuring a character class’s “power” or any such thing, so calling the new class “too powerful”, or “too good”, etc., is not appropriate).

What’s the solution? Not having any class be this dominant in the arena of fun seems like the obvious advice. But in any case, this is a problem indicator.

If a class can’t provide any kind of fun unconditionally, that might be a problem.

This means that for every kind of fun this class can provide, there is at least one other class that prevents the first class from providing that kind of fun. A player of a character of such a class will always be in danger of having his fun shut down or prevented, depending on what other character classes are present in the game.

Of course, there are degrees of conditionality. In a game system with 11 classes, for instance, we can imagine a class that can provide three kinds of fun, but each of those is curtailed by the presence of one other class (and a different class for each of the three kinds of fun). That’s not so terrible. A player of that class will almost always have some of the kinds of fun available, unless the party also contains all of exactly those three other classes—not a very high chance of that.

On the other hand, in that same game system, if we’ve got a class that can provide three kinds of fun, but the presence of any one of five of the other classes curtails all three kinds of fun, then we’ve got a serious problem. Now the party has to be totally devoid of any of those five classes, or else that character’s fun is entirely prevented.

It’s probably not possible for every class to provide every kind of fun unconditionally.

This is a somewhat disheartening conclusion, and my instinct as a GM and as a designer is to resist it. But it seems like an inevitable consequence of the fact that some players enjoy being the best at something. There can’t be two bests.

There’s a caveat, though. A game system can get around that problem by simply not providing the “being the best at something” kind of fun at all, or, more generally, by limiting how many kinds of fun are available. Such a severely limited system could indeed have every class providing every one of the few available kinds of fun…

… and would probably be indistinguishable (or nearly so) from a system where none of the available kinds of fun have anything to do with class selection. That's… one way to solve the “class balance” problem, to be sure. It’s not the way I’d choose.

It's probably possible for every class to provide one kind of fun unconditionally, but such a game system might not be very satisfying.

Think about World of Warcraft (vanilla WoW, before all those newfangled crusades, wraths, and pandas). We could say that there are three kinds of fun: the fun of tanking, the fun of DPSing, and the fun of healing. The game contains the following classes: warrior, paladin, hunter, rogue, druid, mage, priest, warlock.

Well, the warrior provides the first kind of fun (tanking), unconditionally. The mage, warlock, rogue, and hunter provide the second kind of fun (DPSing), unconditionally (everyone can have fun doing damage, at least as long as no one’s comparing themselves to other people—and posting damage meters to raid chat is minus 50 DKP). And the druid, priest, and paladin provide the third kind of fun (healing), unconditionally.

In other words, every class can provide one, and only one, kind of fun, because every class can do one, and only one, thing.

But there’s a reason that this state of affairs didn’t last in WoW. Even before World of Warcraft’s first expansion came out, Blizzard was already changing the game’s mechanics to give various classes the ability to fulfill more than one role. Players simply aren’t satisfied with just one kind of fun; and they don’t want to have to switch characters to experience a different sort of fun.

Of course, those changes brought their own problems. During the Burning Crusade expansion, a paladin could certainly have fun tanking… but only as long as no warriors were around (as the warrior class was better suited to tanking in serious encounters). The paladin class now provided the fun of tanking—but only conditionally. Still, this was an improvement (at least, judging by the clear correlation between changes of this sort and World of Warcraft’s surging popularity, as well as feedback from players.)

* * *

It seems like the next step would be to make some positive recommendations. “In the ideal game system, …”—and then I’d talk about what the Fun Matrix in that ideal system would look like: how the ability to provide the various kinds of fun ought to be distributed among the classes, etc. But I hesitate to do this, without some real-world examples. I’d like to build some Fun Matrices for some actual game systems before I move from the tentative “here are some potential problem indicators” to the certain “here's How to Do It Right”.

XI. The Catch

The Fun Matrix model depends on being able to enumerate the kinds of fun provided by the character classes in the game system.

How the heck do we do that?

I have some ideas about that, which I'll talk about in a future post. For now, though, let’s ask a different question:

What happens if we get it wrong?

That is, what if we come up with a list of the kinds of fun that the classes can provide, and it’s wrong? And, actually, what does it mean for the list to be wrong?

I said, in the very beginning of this post, that there are many taxonomies that people have come up with, of kinds of fun in games. Is one of them right, and the others wrong? Are there multiple right ones? Are there degrees of rightness and wrongness, here?

I don’t think we need to concern ourselves very much with trying to figure out which of the ways to classify kinds of fun is the most correct one. Most of them are probably at least partly “right”, in the sense that each taxonomy gets at some aspect of the play experience. They’re different ways to slice up reality—not necessarily contradictory at all.

And the beauty of the Fun Matrix model is that we don’t necessarily need to choose just one taxonomy of fun. We can construct a Fun Matrix with each taxonomy, and see what insights we get out of it. If the analysis tells us something totally nonsensical—something that does not accord at all with our experience of the game—we discard it. The only criterion that a taxonomy of fun must satisfy, in order to be usable with the Fun Matrix model, is that at least some of the enumerated kinds of fun must be dependent on the properties of the character classes in the system (as described in section II).

So picking the “wrong” taxonomy of fun isn’t something to worry too much about. No, the real danger is that our chosen (or constructed) taxonomy will be incomplete. If we decide on a classification scheme—a way of delineating fun into kinds—but then, when actually listing the categories of fun in that classification, we omit some that should be present, this will distort our analysis. The conclusions we reach may seem sensible, but will be unreliable.

Consider our toy example—the “fighter, shoemaker, paladin” game system. We listed two kinds of fun: the fun that has to do with doing a lot of damage in combat, and the fun that has to do with running a business. Suppose now that we think this is a complete list; as far as we can tell, these are the only two kinds of fun provided by this game system that have anything at all to do with character class. (The game may provide other kinds of fun, we admit—the fun of socializing with your friends, say, or the fun of participating in an epic story—but none that are dependent on class.)

We then build our Fun Matrix, run our analysis (as shown at the end of section VIII), and examine it (keeping in mind the questions and considerations in sections VIII through X). Well, it seems like the fighter is a loser of a class. It’s strictly worse than the paladin; there isn’t any kind of fun at all that it provides unconditionally. If we remove the fighter, we’ll have a perfectly fine little game system: the paladin fights, the shoemaker makes shoes, everything’s just peachy. As it is said, so it is done: the fighter gets the boot.

Then we get angry letters from our legions of formerly-adoring fans. They point out that our modified game system no longer offers players any way to play a character who is an amoral soldier, or a mercenary who fights for any cause as long as the money’s right, or an evil warlord. (The paladin, after all, is always Lawful Good.) It seems that quite a few people find this sort of thing fun! And we took all of that away from them. Whoops.

(Removing the paladin instead of the fighter would also have solved the problem we perceived—and would, of course, have brought with it a symmetric unintended consequence.)

In short: we came up with a taxonomy of fun; we made what we thought was a complete list of kinds of fun in that taxonomy; but we were wrong, because that list was, in fact, missing at least one member (the fun of playing a character who’s a morally questionable warrior). The resulting analysis of our game system was seriously distorted, and led us to make a design decision which resulted in the players having a lot less fun.

Whatever taxonomy of fun we choose to use for our analysis, must be complete. I have a strong suspicion that many, many attempts to analyze class balance in RPGs, and many arguments on this topic, go wrong when they make exactly this mistake. (If I got no other insight at all from writing the preceding several thousand words of this post, I’d consider this one to be worth the trouble.)

In future posts in this series, I'll take a crack at figuring out how to use the Fun Matrix model to actually do some useful analysis of real-world RPG systems. Stay tuned.