Exception-based design isn't revolutionary

[Absentminded_Wizard]

Since Lago started a thread about the effects of exception-based design (EBD) on 4e, I figured I would post some thoughts I've had about the subject lately.

It seems that one problem with discussions about EBD is that a lot of people don't know what it is. Well, actually, such complete ignorance is more common on places like the WotC boards than it is here. However, even on the Den, I think EBD is a lot like pornography. We know it when we see it, but there can be confusion if we actually discuss its properties. So here's my attempt to get a concrete definition for EBD. I'm going to have to work my way through a couple of concepts before my final definition makes sense, so here we go.

Centralization vs. Decentralization

In a completely centralized game, all the rules are presented in one place. If any elements of the game (e.g. cards, pieces, powers) have different rules from other elements, the individual descriptions consist only of references to the central rules list. In a completely decentralized game, all the rules are contained in the description of individual game elements.

Of course, very few games are going to be completely centralized or decentralized. Some traditional games are completely centralized (like Monopoly; no special rules for individual pieces), but I can't think of one completely decentralized game. There's a continuum of centralization, and the games we discuss on the Den are all going to fall somewhere between the absolute extremes.

Rules Dominance vs. Exception Dominance

Rules dominance means that general rules always trump exceptions. In practice, this principle means there are no exceptions. If your game is rules dominant, then any designer who writes exceptions into a supplement should be fired for smoking crack.

Exception dominance means that individual game elements are allowed to violate the general rules. In an exception-dominant game, a rule in a game element's description always overrides the general rule if the two come into conflict.

Unlike centralization vs. decentralization, rules dominance vs. exception dominance is an on-off switch. There is no middle ground.

---

Based on the concepts I've detailed above, my definition of EBD is this:

An exception-based design game is one which is both highly decentralized and exception dominant.

Both parts of this definition must be kept in mind for rational discussion of EBD to happen. For example, if you think exception dominance is the only property of EBD, you will wonder why they couldn't fit as many classes into the 4e PHB as they did in the 3.5 version.

I've also come to an interesting conclusion from this definition. Basically, EBD isn't new at all. It's not something Mike Mearls dreamed up one day recently. It's not even something Richard Garfield dreamed up in the 90s. By this definition, chess is an EBD game. The rules of chess are usually presented in a decentralized way (90% in individual piece descriptions). Chess also has general rules that are violated (knights jump over other pieces, pawns capture in a different direction from their normal movement). People have been doing EBD for centuries, though we haven't given it a name until the last couple of decades.

[zeruslord]

A Thousand Blank White Cards comes pretty close to entirely decentralized, and is purely exception dominated.

You might consider a rules complexity minimum, because the problems with EBD are mostly caused by interactions between pieces.

[Absentminded_Wizard]

If you're talking about 4e's conflicts about action-point rules and dazed condition vs. free move actions, I think the problem is designer laziness. The 4e designers could have worded those powers in a way that didn't have the crazy implications that result. The problem is that WotC seems to think EBD means that designers and devs don't have to think things through or write things out.

[Koumei]

The problem is that WotC seems to think EBD means that designers and devs don't have to think things through or write things out.

This is in fact how they think - such as writing that lazy "Get out of jail free card" about bags of rats.

[Username17]

Some traditional games are completely centralized (like Monopoly; no special rules for individual pieces), but I can't think of one completely decentralized game.

Fluxx. All rules are contained on cards in the deck. The only externalish rule is that one of the cards starts the game in play, and even that card actually has that rule on it.

-Username17

[Manxome]

Rules Dominance vs. Exception Dominance

Rules dominance means that general rules always trump exceptions. In practice, this principle means there are no exceptions.

No, that means there are necessarily no exceptions, not just none in practice. If the general rule is followed even when there's an "exception" that says you shouldn't, then that's obviously not an actual exception.

"Rules dominance," as you have defined it, is completely insane. I can't imagine any game seriously embracing that philosophy. More specific rules always override more general ones, because the only alternative is for specific rules to do absolutely nothing at all, in which case you don't need to have them in the first place, in which case your theoretical rules for how to resolve specific/general conflicts don't even matter because there are no such conflicts.

I mean, consider this rule set:
1. All units attack with 1 die.
2. Red units attack with 2 dice.
3. Follow rule 1 even when another rule contradicts it.

[Kaelik]

What about nobilis? That's, um. Well, at the beginning it's centralized and after a bit it becomes totally decentralized.

[MartinHarper]

Would you regard Chance/Community Cards in Monopoly as decentralised elements? How about the statistics on properties indicating how much rent they take?

[Murtak]

Would you regard Chance/Community Cards in Monopoly as decentralised elements? How about the statistics on properties indicating how much rent they take?

Cards, yes. Rent costs, no.

I think about this is in programming terms, and that means there is a general mechanic (or function or method) called "pay rent" which is called whenever you enter a property. This mechanic is then handed the rent amount as a parameter but the important part is that the pay rent mechanic is defined in a single place.

The chance cards on the other hand have a separate mechanic printed on them, thus are decentralized. If the card in question also features it's own version of an existing mechanic (like "pay 100$ rent for each player who owns a railroad") it is also an exception.

[RandomCasualty2]

Unlike centralization vs. decentralization, rules dominance vs. exception dominance is an on-off switch. There is no middle ground.

Every game is exception dominated. A specific rule always trumps a general one. Like in checkers, where normally you can only move your pieces towards the opposite end of the board, but if you get a king, then you can go both ways. That's an exception to the normal rules.

Especially in any game designed to be extensible, like an RPG, you need to have new rules. And that requires that you have exceptions take priority, otherwise you can't even add anything.

And almost all RPGs are exception based design for the most part. The only one that I've seen try to get away from that is GURPS, but that's because all the base rules are built into the game from the start, and the game is fucking huge and complicated. And even GURPS adds some new abilities in their sourcebooks.

Now, there's the saying "if you have too many exceptions to the rule, you don't really have a rule." which certainly applies to excessive EBD, which I think may be the case for 4E. But the problem with 4E isn't so much that it used exceptions, it's more the case that the designers didn't set up any system to sync up with each other. So you've got the barbarian with an ability that's the same name as a fighter paragon path, or the many different "Evil Eye" abilities of the cyclops. Which is just confusing as hell, because you're attaching the same name to different abilities. But that's not really a symptom of EBD, it's a symptom of bad game design.

[Absentminded_Wizard]

Especially in any game designed to be extensible, like an RPG, you need to have new rules. And that requires that you have exceptions take priority, otherwise you can't even add anything.

Yep. It's pretty much impossible for the kinds of games we talk about here to be rules dominant. The only ones I can think of are some old-school board games. I have to take back Monopoly because of the Chance and Community Chest cards, so the best example I can think of is Sorry. All pieces follow the same movement rules, which are centrally defined.

So it looks like it's possible to be highly centralized and exception-dominant, but impossible to be highly decentralized (or even marginally decentralized) and rules dominant.

[Manxome]

No, it's possible to have exactly one layer of rules so that it is impossible to tell whether you are "exception-dominant" or "rules-dominant" because it does not matter (the trivial case). It is not possible to actually be "rules-dominant" in any nontrivial way, as you have defined it.

Not just "for the kinds of games we talk about here." Ever. In principle.

It's a pointless distinction.

Sorry is not rules-dominant. It has a rule that says that pieces can only move forward, and then special cards that allow pieces to move backwards. It's got a rule that says you draw one card each turn, and then cards that let you draw another card. You could possibly organize the rules in such a way that the general rules are never stated and absolutely everything is on the cards, but that's "both" rules-dominant and exception-dominant, trivially, because no exceptions actually exist, and a strong argument could be made that that would be decentralized, not centralized.

If you honestly believe that "rules-dominant" is something that actually exists in real games, then don't just name random games, identify an actual case of an "exception" being overridden by a general rule. If you can't find such an example, then perhaps some very simple games are trivially "rules-dominant," but as that would also mean that every single game ever is exception-dominant (including the trivially rules-dominant ones), that's a point that is totally unworthy of consideration.

[RandomCasualty2]

No, it's possible to have exactly one layer of rules so that it is impossible to tell whether you are "exception-dominant" or "rules-dominant" because it does not matter (the trivial case). It is not possible to actually be "rules-dominant" in any nontrivial way, as you have defined it.

Yeah, pretty much every game must be exception dominant. To be rules dominant in the given definition basically just means that there aren't exceptions.

And any closed system can be written to be exception free, you just divide the rules up into groups. For instance in Checkers.

Exception Based) A peice must move towards the opponent's side with each move. EXCEPTION: Kings can move in any direction.

Exception free) Rules for normal peices: A normal piece must move towards the opponent's side of the board with each move.
Rules for Kings: A king can move in any diagonal direction.

It's all a matter of semantics. In chess, exception free makes more sense since you have lots of separate pieces and there is no standard, but the exception based writing may be easier for checkers.

But regardless there will be no such system where rules dominate exceptions. And the reason is simple: for an exception to matter, it must override the rule. If an exception cannot do that, then it is effectively unreachable within the game and you might as well X it out.

[MartinHarper]

If you honestly believe that "rules-dominant" is something that actually exists in real games, then don't just name random games, identify an actual case of an "exception" being overridden by a general rule.

Law: the European Convention on Human Rights contains very general rules that override more specific rules in case of conflict.
Dungeons and dragons: "The DM is always right" is a very general rule that overrides more specific rules in case of conflict.

Certainly no game says that the more general rule always overrides the more specific rule. However, it is possible to have games where a few general rules are privileged and override more specific unprivileged rules.

Edit: I just realised that I failed to meet Manxome's challenge. Oops.

[Manxome]

Certainly no game says that the more general rule always overrides the more specific rule.

Then no game is rules-dominant by the given definition:

Rules dominance means that general rules always trump exceptions.
..
Unlike centralization vs. decentralization, rules dominance vs. exception dominance is an on-off switch. There is no middle ground.

Sure, you can have some specialized rules that specifically give themselves priority over other rules. That is not really a case of a general rule trumping a specific one, though, that's a case of a hyperspecific rule ("rule X always wins in the case of a conflict") overriding a less specific one. And it's certainly not a case of "rules dominance" as Absentminded_Wizard defined it.

[MartinHarper]

Well, I'm not ecstatic about AW's choice of "rules dominance" to mean contradiction free, but there you are. Also, there are alternatives to contradiction free rules and specific-overrides-general.

That is not really a case of a general rule trumping a specific one, though, that's a case of a hyperspecific rule ("rule X always wins in the case of a conflict") overriding a less specific one.

Rule A: The DM is always right.
Rule B: The purpose of the game is to have fun.
Rule C: Lettered rules always override numbered rules.
...
Rule 24153: Vigor is a level one Cleric spell of type Conjuration (healing).

I don't understand how you would regard rules A-C as "hyper-specific". They look pretty general to me. They each apply to way more in-game situations than rule 24153. To me, that makes them more general. This is one problem with "the more specific rule wins": it's subject to disagreement over which rule is more specific. Consider:

1. Edged weapons do d6 damage to humans.
2. Axes do d12 damage to creatures.

Which is the rule, and which is the exception? What if I phrase it differently?

1. Creatures take d12 damage from axes.
2. Humans take d6 damage from edged weapons.

[Absentminded_Wizard]

No, it's possible to have exactly one layer of rules so that it is impossible to tell whether you are "exception-dominant" or "rules-dominant" because it does not matter (the trivial case). It is not possible to actually be "rules-dominant" in any nontrivial way, as you have defined it.

Not just "for the kinds of games we talk about here." Ever. In principle.

It's a pointless distinction.

You know, before I respond to this, I'm going to have to take by my agreement with RC's statement that all decentralized games must be exception-dominant. On further reflection, I was way too hasty.

Actually, the distinction isn't pointless, and it is possible to be "rules dominant" in a non-trivial way. What if you have a card game that consists of a small number of central rules and special rules for each card. Card rules may either reference abilities defined in the central rules or create new abilities, but nothing on a card is allowed to contradict a central rule. This hypothetical game meets my definintion of "rules dominant" and is still decentralized and has specific rules. These questions of dominance only come into play if the specific and the general conflict. If all specific rules are unique and don't intersect the general rules at all, a "rules dominant" game can have more than one layer of rules.

Now, it's true that the above system shouldn't have any exceptions to the general rules ever printed on the cards. The "rules dominance" clause only comes into play if somebody doesn't get the memo and writes an exception on a card that somehow makes its way into the game.

Sorry is not rules-dominant. It has a rule that says that pieces can only move forward, and then special cards that allow pieces to move backwards. It's got a rule that says you draw one card each turn, and then cards that let you draw another card.

It's been so long since I played Sorry that I'd forgotten about the backwards cards.

You could possibly organize the rules in such a way that the general rules are never stated and absolutely everything is on the cards, but that's "both" rules-dominant and exception-dominant, trivially, because no exceptions actually exist, and a strong argument could be made that that would be decentralized, not centralized.

If absolutely everything is on the cards, the game would be decentralized. If the general rules and the card rules never contradict each other, the question of rule vs. exception dominance never comes up.

If you honestly believe that "rules-dominant" is something that actually exists in real games, then don't just name random games, identify an actual case of an "exception" being overridden by a general rule. If you can't find such an example, then perhaps some very simple games are trivially "rules-dominant," but as that would also mean that every single game ever is exception-dominant (including the trivially rules-dominant ones), that's a point that is totally unworthy of consideration.

I'm not sure if it exists in an actual game, but it's not impossible.


MartinHarper: I think I might have to reconsider how much of an on-off switch rules vs. exception dominance is. It looks like there's room for middle ground after all.

In chess, exception free makes more sense since you have lots of separate pieces and there is no standard, but the exception based writing may be easier for checkers.

You could write the moves for chess pieces in a totally decentralized way that gets rid of exceptions like "no piece may jump over other pieces, except for the knight." There is one general rule that's hard to get rid of, though: the rule that each player may move only one piece on his or her turn. This rule also has an exception: castling.

I guess you could write the rules to say something like, "On your turn, you may make one move" and then have a list of possible actions like "move 1 bishop any number of unoccupied squares in any diagonal direction." That organization would allow castling to be just one more rule, but it would be the most cumbersome thing to read.

[Psychic Robot]

1. Edged weapons do d6 damage to humans.
2. Axes do d12 damage to creatures.

FFFFFUUUUUUUUU--

[Manxome]

Rule A: The DM is always right.
Rule B: The purpose of the game is to have fun.
Rule C: Lettered rules always override numbered rules.
...
Rule 24153: Vigor is a level one Cleric spell of type Conjuration (healing).

I don't understand how you would regard rules A-C as "hyper-specific". They look pretty general to me. They each apply to way more in-game situations than rule 24153. To me, that makes them more general. This is one problem with "the more specific rule wins": it's subject to disagreement over which rule is more specific.

Rules A and B are quite general. Rule C only applies to situations where multiple other rules overlap, which seems like a very highly constrained range of applicability to me.

You are correct that rule specificity isn't necessarily totally ordered, but that doesn't change the fact that more specific rules win when you can tell which rule is more specific.

And I guess I was somewhat overbroad. If you have two rules that give contrary instructions in some special case, and you add a new rule that says how to resolve that special case, that new rule is more specific than either of the previous rules, and so it can win on ground of specificity. If you instead create a rule that gives higher priority to one particular rule or set of rules, it means that you eliminated the contradiction and so rule specificity doesn't need to be considered. It's a meta-rule, in that it tells you how to apply rules rather than directly telling you how to play the game, but it's still a rule whose existence prevents the occurrence of a contradiction.

Also, "the DM is always right" is not an actual game rule, it's a colloquial admonition to let the DM guide the game. Taken as an actual letter-of-the-law rule, it's an incredibly sucky rule, not to mention that it's trumped by higher authority (no matter what you print in your rulebook, you can't change the rules that govern the reality in which your players live, like the one that says 2+2=4 whether the DM agrees or not). And heaven forbid your DM utters the words "oops, I was wrong."

[Murtak]

So it looks like, technically speaking, pretty much every game uses EBD, according to the original post. Here is an example:

Design intent:
We want there to be a spell that makes people sleep, but elves should be immune to it.

Horrible design:
The Sleep spell has an exception it it, where elves are immune.

Bad design:
The elf race has an ability that makes them immune to the Sleep spell.

Good design:
The elf race has an ability that makes them immune to spells with the sleep keyword. The Sleep spell has the sleep keyword.

- The first example runs into trouble when you add more sleep immune races (now you have to change every spell that applies) or when you add new sleep spells (now you have to remember to add the elf exception to any new spell).
- The second example deals fine with adding more races, but still has issues when you add more spells.
- The third example deals fine with both, assuming the list of keywords is not too long and keywords are sensible.

All three examples however use exceptions, all are exception-dominant and all are decentralized. I don't think AW's categories are very helpful here. We have to ask ourselves "what is the problem with EBD anyways?". And the answer is "it leads to rules fragmentation and duplication, which in turn leads to an unmaintainable mess that is hard to write new rules for".

So the keyword is rules duplication. The first exception is always easy to make. It is the first instance of a new rule and no matter where you put it, it is more text to remember. The question is, what happens when you add more stuff that triggers the exception (or is supposed to trigger the exception) or what happens when you add similar exceptions. Ideally you will not have to change any existing rules at all, and you will at most have to reference the rules you wish to apply to the new stuff you are adding.

In other words: don't repeat yourself (thus DRY-Principle).

[JonSetanta]

Consider:

1. Edged weapons do d6 damage to humans.
2. Axes do d12 damage to creatures.

Which is the rule, and which is the exception? What if I phrase it differently?

1. Creatures take d12 damage from axes.
2. Humans take d6 damage from edged weapons.

(head esplode)

The result for the first instance is a grand sum of d12+d6 when using edged axes, except when there are rules stating "You can't add dice from varied properties of a weapon".

The result for the second instance results in nonhuman creatures taking d12 damage while humans (if creatures) take d12+d6.

If assuming FILO/LIFO then, in order, a gamer would observe rules in the order they are presented and alter the system as each change overrides the previous.

Otherwise you thrown the book down and go play Melee or Brawl.
Or Starcraft.

Good design:
The elf race has an ability that makes them immune to spells with the sleep keyword. The Sleep spell has the sleep keyword.

Also known as 'tags' in other interpretations.
I think D&D was began without such concepts many years ago and as a result required (or still requires) a makeover to use tags/keywords.
While 4e did have an opportunity to do so, I'm stunned by the actual results.

[Draco_Argentum]

EBD is like poor programming where code is repeatedly written in where needed rather than being done in a reusable manner. Good coding uses objects and functions to keep code duplication low. It allows maintainability and comprehensibility by having stuff done once.

In essence I agree with Murtak.

[MartinHarper]

A contradiction-free approach would be:
Slumber: puts people to sleep, except if they are immune to sleep spells.
Elves: immune to sleep spells.

EBD is like poor programming where code is repeatedly written in where needed rather than being done in a reusable manner.

In that case, Rules Duplication like an inlining compiler. The design file for a M:tG card just says "Cycling 2", and when it is compiled it is expanded into "Cycling 2 (Pay 2: discard Celtic Ninja and draw another card)". The design file for a Kobold (hopefully) says "Shifty", and when it is compiled it is expanded into "Shifty : minor : Kobold Ninja shifts one square".

[Absentminded_Wizard]

A contradiction-free approach would be:
Slumber: puts people to sleep, except if they are immune to sleep spells.
Elves: immune to sleep spells.

3.5 did this with one magic item vs. DR. I believe they basically wrote the rules for DR in the DMG (though not the SRD, with hilarious results) as "DR x/- ignores x points of damage from any weapon that doesn't ignore DR." Then the item description said, "This item ignores DR."

[Murtak]

A contradiction-free approach would be:
Slumber: puts people to sleep, except if they are immune to sleep spells.
Elves: immune to sleep spells.

What is the point of wording rules like that? Now you have an extra phrase you have to tag onto every rule that might ever have an exception. I'm not seeing any improvement over:
Slumber [Sleep]: Puts people to sleep.
Elves: Immune to sleep effects.

In that case, Rules Duplication like an inlining compiler. The design file for a M:tG card just says "Cycling 2", and when it is compiled it is expanded into "Cycling 2 (Pay 2: discard Celtic Ninja and draw another card)". The design file for a Kobold (hopefully) says "Shifty", and when it is compiled it is expanded into "Shifty : minor : Kobold Ninja shifts one square".

That is one way to handle things, right. Feats work like this. This approach makes no sense when you have unique or close-to-unique abilities which should still interact with other abilities and rules however (say, a ghoul's paralysis).

3rd edition worked pretty well in this regard. You have keywords (like the spell descriptors), types and categories (spell schools, creature types), base mechanics which are properly referenced (AoOs) and feats to abstract commonly used mechanics into a neat bundle. Sure, a lot of stuff in 3rd edition is horribly implemented and balanced, but the foundations are pretty damn sound.