The “Four” Forms of Game Design

In the near future, I’m going to be publishing more articles about game design.  I wanted to start by picking apart some of the methodology of ClockWork Game design by Keith Burgun, one of my favorite Game design personalities.  Firstly I want to thank Keith, I had lost interest with game design for a long time, and his podcasts and articles got me interested in it again, I think he’s doing something incredibly valuable, and I wish that a lot more game designers were willing to be more like him.  With a field as young as game design is, I honestly think it’s far better to put your opinion out there and be wrong than it is to stay quiet and pretend you know everything.  With that said, Keith is somewhat infamous as being “Opinionated”, and while it would be easy to look at why people find him so polarizing, he’s said some fairly judgemental things about other games and game developers, I think it’s far more valuable to look at his work and try to tease open the methodological problems that cause him to say things that rub many people the wrong way.

One of the cornerstones of Keith’s method is his theory of the four forms of interactive media.  Keith talks a lot about his dissatisfaction with the way we define terms in game design, and in part to combat that and in part to stake a claim on the term “game”, Keith has introduced four categories by which we can define any interactive system (which colloquially we would call a “game”).

In his own words, he summarizes the four forms in this video:

For clarity sake or for those who can’t open a video I’ll re-summarize the forms here before I continue.


Toys are bare interactive systems.  Toys don’t have any prescribed goal, therefore all you really do with them is figure out the limitations for how they work.  Players “play with” or “explore” a toy.  The primary value they provide could be called “mapping”.  Minecraft or a football is a toy.


A puzzle is a bare interactive system + a prescribed goal.  Players attempt to manipulate the system to reach the goal state.  The primary value is “solving”.  A portal level or a Rubix cube is a puzzle


A contest is a bare interactive system + a prescribed goal + a session limitation.  Contests are primarily about measuring.  How fast can the player reach the goal state?  How many goal points can a player score?  It’s not good enough to solve the problem, a player is trying to maximize some performance metric.  Good contests are not solvable, the player should always be able to score higher or run faster no matter how good their last run was.  Marathons, bowling, weight lifting are contests.


Games are contests + obfuscation of the game state.  This obfuscation forces players to build and rely on small bits of understanding to hopefully inch towards the goal.  The primary value is “understanding”, a good game has many bits of “understanding” and we call that “depth”.  Chess and Go are games.

For the most part, I think there’s value in these forms, I think there is useful taxonomic information here, but I believe there are very serious problems with the way Keith uses them.  I also believe this taxonomy is not complete, although I won’t attempt a comprehensive alternative taxonomy in this article.  Anyways, let’s get started.

Let’s talk about the “Game” form

In one of his podcasts, Keith claims that he has developed this form to provide proper definition to the word “game” in contrast to a solid and sensible definition of the word “puzzle” he had encountered previously.  If you’re like me, from the definition given so far, it’s fairly difficult to picture what Keith even means when he says a “game”.  Why is chess a game and not a puzzle? You can see the entire game state and you have only one clear goal, to win.  You don’t really score a higher chess, you just win.  I think what happened is Keith originally had a definition somewhere along the lines of “a game is a competition against another player”, but wanted to justify the existence of single player games, so he realized that while you can see the whole board state in a game like chess, you can never see the inside of the opponent’s head.  The lack of an ability to know what move the next player will make mean most of the “game state” is hidden from the player, they can’t ever know to react to a move another player hasn’t yet made.  Therefore, all you really need to do to make a single player “game” is to hide enough information that the player can’t solve the game, similar to how a player can not “solve” chess (sure, computers can, but no can accurately be said to have solved chess).

What about the part where you don’t win chess points?  I imagine the argument Keith makes is that with a traditional puzzle it doesn’t really matter how well you solve the puzzle.  If you finish a jigsaw puzzle you have a picture, if you don’t finish it you don’t.  Victory is binary.  Where as in a multiplayer game if you play poorly and the other player plays well, you loose.  The quality of your mastery of the game affects your ability to consistently reach the binary win condition.  Therefore it’s not good enough to be able to win a single round of one of Keith’s games, the value comes out of repeatedly being able to win against more and more difficult scenarios or opponents.

Why should we care about the design forms?

In my opinion, the reason Keith created this taxonomy is his primary philosophy behind game design is that you should design your game to have a primary “goal”, and every other element of your game design should be crafted to support this goal.  Therefore, if we can distill the reason a person is interacting with our system of rules into one of these four forms, we can craft our rules to as exactly as possible support the person’s pursuit of that goal.  To Keith, we can set the ultimate binary goal for our game, and we will know the player is pursuing “understanding”, so we distill out everything that doesn’t support understanding or winning via the win condition more interesting, and we should ultimately get the purest, and therefore the best possible game (provided we also apply some other rules for making good games).

I’ve listened to more than half of Keith’s podcasts, and I’m working my way through the rest.  In one of his podcasts, he says something along the lines of “You cannot have more than one primary goal, that couldn’t even work”.  When Keith talks about the four forms, or other games, he often throws around statements like “Minecraft is a toy” or “League of Legends is a game”.  Keith also has a lot of prescriptive rules like “grinding kind of works for a toy, but it’s absolutely unacceptable for a strategy game”.  I think something poisonous has gone on here that has caused Keith to inappropriately throw out a lot of useful design tools in the misguided pursuit of form purity.


Before I say anything else, I do want to recognize that I find Keith’s linguistic sin of redefining commonly used and well-established words appropriately frustrating.  I find it especially misguided because Keith has chosen some of the most important items in the shared namespace game developers rely on as his candidates for redefinition.  Keith, it’s common social convention, pick a new word, don’t try to squat on words we all use and already have polarizingly well-established definitions.  You’re giving the trolls fuel for the fire in their attempt to argue Minecraft players aren’t real gamers, and there’s really no good reason for it.

The huge problem, in my opinion, is that it’s not even actually possible for a game designer to pursue a pure form, and because it’s an impossible goal to achieve, there’s can be no value in a system of rules which is “pure”.  By “impossible”, I don’t just mean that it’s inconceivably difficult, I mean it’s actually by definition a logical contradiction.  By definition, a puzzle is effectively a toy plus a binary objective.  By definition, a competition is a puzzle plus a measurement of goal accomplishment.  By definition, a “game” is a competition with “obfustication”.  Therefore, I cannot have a game without there being a toy buried under the hood somewhere.  Not only that, but like there is an absolutely essential toy buried somewhere deep under the hood of chess, most good toys also have interesting “games” buried under their hoods as well.

What is a game?

It’s easiest to understand what I mean with examples.  In Starcraft II 1v1, your goal is to get into the state where it’s impossible for the other player to prevent you from destroying every unit and building they have.  But when you start your first 1v1 game of Starcraft II, while you may be aware of what the game tells you your goal is, you do not know what happens when you move your units around, you don’t know what happens when you press various buttons, you have some information from the theme of the game, but fundamentally you do not have a map of Starcraft II.  Whether you as a designer like it or not, every player playing Starcraft II for the first time must map the system, and therefore they start in a toy.  This toy quickly evolves into series of small puzzles.  How do I build a marine?  How do I get a battle cruiser?  These by definition are puzzles, because there’s an interactive system you’re interfacing with that has a binary goal state.  These puzzles evolve into competitions, because eventually, it’s not good enough to build a marine, you need to build a marine as fast as possible.  You start having to improve mechanical skills in pursuit of optimizing goal state achievement.  You could freeze any given subset of rules in the Starcraft II system and it will either be a toy, a puzzle, a competition, or even it’s own fully formed game.

What is a Toy?

“Toy” is the other fairly problematic category, because it’s impossible to map a bare interactive system without at very least performing binary objectives on subsets of that system.  There is no prescribed objective in Minecraft, therefore it is a toy.  But how do you build a zombie-proof house?  “I want to build a zombie-proof house” is a statement describing a binary objective, so there must be a sub-puzzle in Minecraft where you build zombie proof houses.  In order to map the possibilities of Minecraft, you must interact with subsets of the Minecraft rule set which are puzzles.  In reality, you also must interact with subsets of the Minecraft rule set which are competitions “How fast can I mine enough iron for an anvil?” and games “Kill the other player before he kills you”.  So while a bare interactive system with no goal may appear to be the most basic type of interactive system which can be built, it’s not in fact possible to build an interesting toy which does not require the player to either interact with subsystems which are puzzles, competitions, and games, or alternatively to expand the toy system to use it as a subsystem for their own puzzle, competition, or game.


What does this tell us about game design?  In a nutshell, you are not designing a game about completing your primary objective.  You’re designing layers and layers of small games which combine into a game about completing your final objective.  I’d argue that Keith is right that any good game only has one primary objective, but there are so many absolutely essential sub-objectives and component games that must be played along the way to reaching that objective that “purity” of form is an absolutely absurd criterion for quality in game design.

It absolutely is possible to combine subsystems in such a way that you corrupt the pursuit of your primary objective, but that’s not because you’re inappropriately introducing impure components to your design, it’s usually because you’re setting up essential sub-objectives in such a way that the conflict with each other.  You set up a competition the player would like to optimize, then invalidate it’s results half the time with a dice roll, or you set up a puzzle a player has to solve to play optimally, but don’t give them enough time to solve the puzzle.

Saying “Minecraft is a toy” doesn’t imply that the human brain can quickly dig in and understand what it’s about.  Making a good game has far more to do with understanding the sub-toys, sub-puzzles, sub-competitions and sub-games you’re asking the player to play far more than it has to do with pursuing purity.  So it’s nonsensical to make a statement like “grinding is only good for toy’s but it can’t work for games”.  There’s a vital toy within that game and if that toy is stronger, and there’re no other subsystem conflicts, then the game is stronger too.  Any design pattern that can work for any form can work for any other form, because the strengthening of an essential sub-component of the parent, provided it doesn’t conflict with other sub-components, must also strengthen the parent.  This doesn’t mean as a designer you must implement or like grinding, but it does mean that the value of grinding has nothing to do with whether you’re making a game, a puzzle, or a toy.

In a nutshell

I think the four forms are VERY valuable tools for thinking about your game and how it works.  That being said, I think they need a lot of work, because I can easily show that a competition, or even a game, can be a sub-component of a toy; but no one has yet described the process by which that happens.  Along those same lines, I think the fact that puzzles and “games” both have binary objectives probably means they are closer to the same thing than the basic four forms model implies.  I don’t like the names Keith has chosen for the forms, I really hope we can come up with better names in the near future.  While I do think we might be able to say things like “don’t add your grinding into the game layer”, I don’t think we can say things like “child toys of a parent game shouldn’t have grinding”, and therefore “games shouldn’t have grinding” also is logically flawed.  I realize this article creates more problems than solutions, but I don’t think the proposed solutions were good enough anyways, so it’s not healthy to get too stuck on them.  I hope we can continue to improve on the forms model in the future with more thought and discussion.


4 thoughts on “The “Four” Forms of Game Design

  1. Hi! It’s me, Keith! Thanks for writing this. Thought I’d address some of your concerns.

    Why is chess a game and not a puzzle? There are specific requirements that each form needs, which I’m sure you know. But beyond that, in a very simple way you can just say: puzzles are designed to be solved. Chess was designed *not* to be solved.

    > is that it’s not even actually possible for a game designer to pursue a pure form

    Huh? I know tons of pure puzzles. How *isn’t* a “put square peg in square hole” children’s toy a pure puzzle? Or even a Professor Layton puzzle? Garry’s Mod seems to be a pure toy. A weightlifting contest seems to be a pure contest. As to games, I agree there aren’t as many examples of a “pure” game but I think Auro is one example. That was partly why I made it, to demonstrate that such a thing could be made.

    >You’re designing layers and layers of small games which combine into a game about completing your final objective.

    The problem with this as a “revelation” (besides the fact that I disagree with it) is that you’re not moving the ball down the field at all. This is already what everyone does. This is standard practice and doesn’t need to be written about, it seems to me.

    1. Amazing to hear from you Keith!

      I suppose it’s pretty fair to accuse me of being reductionist when so far my only post on this blog is pulling back on something you put out there. I promise I’ll put stuff up here in the future which I’m sure I’ll be excited about but will look equally like a shot out of bounds to you as the four forms currently do to me. I honestly tried to steel man your argument, but there’s really is almost a mathematical contradiction to the definitions as you give them right now. I honestly hope you can clarify further and invalidate this whole article. I think it goes without saying my perspective on game design is quite different than yours, I’m mostly interested in making what you would call toys (at least right now) I hope you can understand that your poking sharp sticks at some of the principles I lean on heavily sparked a desire to get off my butt and do something.

      On your arguments: “solved” is a complex word with multiple meanings, when we say “chess isn’t meant to be solved” the “solved” we mean is true of almost all puzzles as well. Puzzles are meant to be compelling in the act of getting to the goal state, it’s an entirely different activity mapping out the optimum move for every puzzle state. Actually, in general we have several steps a long the line with a puzzle where we figure out new ways to optimize getting from zero to solved1. The process of getting to those optimization steps may or may not also be designed to be interesting and fun, a good designer will tweak dials at those levels as well. Many puzzles are quite deep.

      Puzzles, including block in hole, always have a bare interactive system that has to be mapped before you can get to the puzzle. That system is trivially easy to map visually for human adults, but infants actually spend more time mapping the system than they do solving the puzzle. It makes sense to build a puzzle’s bare system to be entertaining to map.

      Garry’s mod and most other games you define as toys are not designed to be fun to map, they’re designed be the cornerstone for players to add a few of their own rules and turn the system into puzzles, competitions, and games. also, Mapping is kind of a weird thing because it’s a binary goal which implies it’s closer to puzzledom. Garry’s mod has a bunch of small puzzles games and competitions built into it as well. Garry’s mod specifically is a digital version of this

      1. >when we say “chess isn’t meant to be solved” the “solved” we mean is true of almost all puzzles as well.

        To clarify the distinction:

        I’m designing a puzzle: I design it so that it’s hard to solve, but I expect someone to solve it in a matter of hours (if not minutes)

        I’m designing a game: I design it so that it’s as impossible to solve as possible – thousands, maybe millions of players playing this for hundreds of hours each and not finding the solution.

        To me that seems like a pretty large distinction.

        I agree that puzzles contain the ingredients of a toy. Also contests contain the ingredients of a puzzle, and games contain the ingredients of a contest. This is part of the hierarchy of forms as I describe it.

        >Mapping is kind of a weird thing because it’s a binary goal which implies it’s closer to puzzledom

        One thing that may help us bridge this gap we’re having is if we could agree that *everything in the universe* is about mapping / solving / whatever. Everything, technically, has that puzzledom-like element to it, if you spend enough time on it.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s