This piece isn’t a guide on how to make a better game, nor a manifesto on what counts as a successful, fun, or great one. Those are different questions, belonging to different layers.
Its only purpose is to stimulate thought and help game developers understand the thing they are creating before they decide what else they want it to become.
Table of contents
Misconceptions and Associations
Sometimes, while developing games on my own, I catch myself returning to the same irritating question. There are plenty of things people call games, and yet something in my intuition keeps insisting that a few of them are actually…less game than they appear to be. So what actually makes a game become itself?
- Graphics, story, world-building? Painting, literature, comics, and film already know what to do with those.
- Flow, pacing, music? A concert, a film, or a stage performance can be constituted through all three.
- Progression, competition, goals? Too generic. Careers, arguments, education, sports, and exercise can organize themselves around them perfectly well.
- Interaction? A light switch is interactive. A DVD menu is interactive. Replacing a play/pause button of a movie with the controller pushing upward did not make that into a game.
- Mechanics? Closer, but not yet. We first have to distinguish mechanics from Rules. A Rule is a pure logical relation. A mechanic is one expression of that relation. One Rule may be expressed through many mechanics, and two things that look nothing alike may still instantiate the same Rule. In Xiangqi, a pawn advancing one square and a cannon moving in a straight line to leap over a piece and capture share no movement mechanic, yet both consume one turn. Notice that this last sentence only became true once I said which comparison I was making. Hold onto that. It turns out to be the whole problem, and it gets settled further down.
None of these things is worthless. They may enter a game, dominate it, or become the entire reason anyone cares about it. But they can already constitute forms and activities whose identities do not depend on gameness. A story does not wait for a controller to become a story. Music does not need a skill tree before it becomes music.
Strip those layers away and they may remain intact as themselves. They do not yet explain what gameness adds.
So, what makes a game… a game?
If it’s none of those, then what the heck is a game?
This is not an attempt to decide which products deserve that label. I am asking something smaller: what must already be present before anything else can be built upon the game?
I am not looking for the answer in a “magic circle”, a “storytelling machine”, “interesting choices”, “meaningful interaction”, or any other phrase assembled from experience, value, or observation. Those may describe what someone does with a game, what someone receives from it, or what someone hopes it will become. They do not yet reach whatever the Game itself cannot do without.
How about Family resemblance?
☝️🤓 ‘But sir, you’re dogmatic, Game has Family resemblance, it may not have common essence!’
Family resemblance observes that things gathered under one concept may share overlapping similarities without one obvious trait running through the entire set. It does not establish that no common condition exists, and the word may cannot do that work for it.
An invariant is a non-trivial condition every Game within a declared domain and representation must satisfy. An essence would additionally explain what makes the thing a Game. A complete definition would decide membership in both directions. An essence must contain or entail some invariant, while an invariant need not rise to the level of an essence.
This inquiry asks for the weaker floor. Rejecting essentialism does not reject that floor, and family resemblance supplies no reason to abandon the search before a candidate has even been attacked. So here is one.
The Necessary Condition
I should state my bias before asking the argument to carry it: I think a game is closer to a simulation than to a film.
That does not mean it must imitate physics, pursue realism, or calculate thousands of little objects. A game may simulate a duel, an economy, a conversation, a legal procedure, a deck of cards, or a world whose laws would be insulted by ordinary gravity. What matters is that there is a configured form capable of continuing according to its own relations. An event occurs because the form admits it and determines what it can lead to, rather than because a finished sequence has already placed it there.
This gives the argument a direction. Begin with a simulation-like form, ask what lets an intervention become structurally operative inside it, then see what that relation forces us to distinguish.
Rules, first
We’ve already separated Mechanics from Rules. The next question is what a Rule must contain before any mechanic can express it.
A Rule distinguishes an applicable situation, the inputs it accepts there, and the results it permits. It may transform something, preserve it, refuse a proposed transition, disclose a reading, or constrain several other Rules at once. It may also operate without receiving any intervention at all. Operation alone therefore says nothing about Agency.
This already tells me that a Rule cannot be built from proper nouns.
A concrete occupant enters through a typed position; the law belongs to the relation between positions. Any difference in how that occupant is treated must be carried by some represented property, status, type, authority, or role. A proper noun is not an undeclared input to the universe.
The same applies when a Rule concerns something unique. Uniqueness may be represented as a property, identity as a value, and exclusive authority as a Role relation. Once represented, the distinction is available to the Rules. Until then, a bare name is merely an exception pretending to be a law.
So a Rule contains typed positions and the relation among them. Configurations later provide occupants and values for those positions. One Rule clause may be applied many times, and several Rule clauses may participate in one transition. The clause, its particular application, and the intervention that may have prompted it are three different things.
That last separation matters. An intervention enters a Rule as input. It does not write the result directly. The Rule decides whether the input is admissible and which consequences may follow. The intervention, its particular application, and its consequences remain different objects even where some of their represented values coincide.
Rules also need scope. A fact existing somewhere in the current configuration does not make it available to every Rule. One clause may carry a distinction into another only through a position their scopes actually share, and the second clause must genuinely depend on that position. The configuration is not a room in which every clause can overhear every other clause.
This is what I mean by a Rule intersection. Two Rules intersect where the output position of one becomes an input position of another in a later application. The intersection opens a route. It does not guarantee that the route carries anything useful: a distinction must survive the relation instead of being ignored, merged, or erased.
The same restriction applies to information. Transmission, validation, proof, disclosure, and knowledge are not five names for a value arriving somewhere.
- A distinction is transmitted when a chain of Rules actually carries it.
- A claim is validated when the Rules establish it from the inputs that support it.
- A proof is a represented certificate that may itself be preserved, transformed, or disclosed.
- A disclosure makes a reading available through a declared interface.
- Information supplied by the form can become knowledge after a Source receives a reading and updates on it. A Source may also arrive with knowledge acquired elsewhere.
One relation may support another, but none appears merely because a message with a reassuring label has arrived. Information travels through Rule intersections, under the same scoping discipline as every other operative difference.
At this point the Rules describe more than a bag of local clauses. Taken together, they determine which complete transitions are admitted. Clauses that apply at the same point must agree on shared results. Anything they leave untouched persists only because persistence is itself part of how the configured form carries the present into the next transition.
That gives us a raw carrier of everything the Rules can currently distinguish and update. Call one complete assignment of those represented quantities a Configuration.
So, what is State?
It is tempting to rename Configuration as State and move on. That would leave too much accidental storage inside the concept.
Two Configurations may differ along represented coordinates that no Rule can ever read. If no declared reading, admitted input, or possible continuation can reveal the difference, the configured form itself has no use for it. Counting it as a different State would let inaccessible representation manufacture structure merely by existing.
So I need a stricter name. Two Configurations belong to the same State when every finite inspection available to the configured form gives the same answer from both.
An inspection may read what is currently disclosed. It may ask which contributions are admitted. It may also look through a possible contribution and inspect what could be read after it, then continue doing so through a finite chain. Because the form may branch internally, an inspection keeps the whole set of supported results. Because several questions about the same branch may be correlated, it keeps those answers together rather than flattening them into unrelated lists.
State is therefore the complete present condition as far as the configured form can ever make a distinction operative. It is neither the whole machine nor a screenshot.
A remembered event belongs to the current State when it can still change a reading, an admission, or a continuation. If the route taken to arrive here can matter later, some represented trace of that route separates the present States. If every future inspection agrees, the different histories have genuinely arrived at the same State for this form.
This also separates State from what one particular participant can see. A distinction belongs to the State whenever it can still alter a reading, admission, or continuation, even where one Role cannot read it. Each Role receives only the disclosures the Rules make available to it. The full State gathers every distinction the form can inspect; an occupant acts through a narrower and possibly uncertain view.
Now take every Configuration reachable from an admitted beginning, every transition among them, and every branch, loop, or continuation the Rules support. Call that entire reachable structure the Possibility Space.
Possibility Space is not a synonym for freedom. It says nothing yet about expression, enjoyment, depth, or whether the available futures are interesting. It is simply the reach of the Rules. Its size does not determine how much of it carries intervention-sensitive difference.
It also contains more than a list of destinations. Two contributions may eventually reach the same place and still open different continuations on the way there. Conversely, two immediately different outcomes may be indistinguishable once everything the form can inspect is considered. What matters is the branching structure the Rules preserve, not the artwork attached to the leaves.
The Possibility Space belongs to the configured form whether or not an Operating Instance is running. Operation introduces another layer.
From the form to one instance
Call one concrete operation of the configured form an Operating Instance. It begins from an admitted Configuration and commits one ordered path through the Possibility Space. Call that committed path a Trajectory.
The Possibility Space may branch. The Trajectory does not branch retroactively. Once one outcome has been actualized, the alternatives remain counterfactual parts of the form, while the instance has acquired one history. Another Operating Instance may actualize another Trajectory without changing the configured form at all.
Between two consecutive States lies a Transition Window. It is a logical position in the Trajectory, not a promise about seconds, frames, animation length, or processor time. A consequence may become inspectable immediately or only after several later transitions. Its structural delay is about how deeply one must look into the continuation before the difference appears.
This distinction prevents several layers from collapsing into one another. Rules define the form. The Possibility Space is what the form admits. An Operating Instance actualizes a Trajectory. The source observing or intervening in that instance is yet another thing.
So the next question is not whether somebody “interacts”. It is how a difference supplied by somebody can enter the Rules.
A Role is a typed position through which contributions may become operative. It determines the contribution values authorized at that position and the information disclosed to it. A Contribution is one value supplied through the Role at a Transition Window. All Role contributions admitted together there form a Contribution Profile. A Source is whatever supplies one or more of them.
A Source need not be a person, singular, aware, or intentional. The Rules need only a supplied value attributable through the relevant binding. An Operating Instance binds the Source to a Role for that window.
Authority still does not produce Agency. Authority determines which contribution values are admitted through a Role; Agency additionally requires variation among those values to separate Continuations. The cardinality of an input alphabet alone says nothing about that relation.
Nor does disclosure produce Agency. Disclosure determines which information is available to a Role; Agency compares Continuations under admitted contribution variation. Neither relation entails the other. Informed control requires both a structural alternative and some policy capable of using the available information. Guessing keeps the first relation and lacks the second.
What continues from here?
After an admitted Contribution Profile, the form may permit several immediate outcomes and many later paths. Take the whole branching future opened by that profile. Call it a Continuation.
Two Continuations count as different when some finite inspection can eventually separate them. The first inspection depth at which that happens is their Latency. Latency orders how far one must look before a fork becomes legible. It is not elapsed time, consequence magnitude, or a ratio in which depth six is somehow twice depth three.
This is why the next transition alone is too weak a comparison. Continuations are separated by finite inspections across the whole supported future, whether or not the distinction is already disclosed at depth one. Agency belongs to the Continuation, not to the amount of movement on the screen.
Internal nondeterminism stays inside one Continuation. If the same complete Contribution Profile can lead to several supported outcomes, the form branches without attributing that branch selection to any Source. Randomness may decide what happens. Agency asks whether changing an admitted profile can change the branching future that is available.
Now Agency
The required comparison is finally visible.
At one reachable State, take two admitted Contribution Profiles. Let them differ only through the Role, or block of Roles, being evaluated. Hold every contribution outside that block fixed. If the two profiles open different Continuations, call the relation Agency.
Holding the rest fixed is the important part. Without it, a difference produced elsewhere can be credited to whichever participant happens to be convenient. With it, the counterfactual is local: this Role block could have supplied another admitted contribution under the same surrounding commitments, and the form would have opened another continuation.
An Agency point is a reachable place where such a counterfactual difference can enter. It belongs on the input side of the Rules. The Rules receive the contribution, determine its admissibility, and carry or erase its distinction through their consequences. A Rule does not need to create an Agency point in order to operate, and no Agency point needs to be the output of a Rule.
Agency here is structural availability. It does not say which contribution was actually chosen, which immediate outcome became actual, or whether the contribution caused that particular outcome in every useful sense of causation. The State and binding may be actual while the alternative contribution remains counterfactual.
Call the Source occupying an Agency-supporting Role block an Agent at that binding and Transition Window.
Agenthood is therefore local to that binding and Transition Window.
Sometimes no single Role can vary while all others stay fixed, yet a block of Roles can vary together. That is still Agency. The minimum number of Role coordinates that must change tells us the Agency arity of that witness. It counts structural coordinates, not people, cooperation, social importance, or emotional intensity.
Different Agency witnesses can also have different shapes. One fork becomes legible quickly; another remains latent for many transitions. For one witness, every immediate outcome under both Contributions may retain a route to one common State under one finite bound; for another, no such bound exists. Call that question Reconvergence.
Reconvergence does not undo Agency. If two Continuations were distinguishable, a later common State does not erase the fork that occurred. It says every immediate outcome retains some admitted route to that State within the shared bound. Possibility is not actuality, and those routes are not a guarantee that an Agent can force them.
Agency, Latency, arity, and Reconvergence remain attached to the same witness. Pulling them apart into free-floating scores would lose the very relation that gave them meaning.
Back to Game
Before applying any bridge to Game, the structural representation must already be fixed. It decides which differences enter as Rule-governed Role contributions, which remain internal branching after every Contribution Profile is fixed, and which are discarded. Call the configured form retained by that declared reading a structural extraction. No canonical extraction is assumed.
A structural extraction is Structurally Playable when at least one reachable State contains an Agency point.
My single bridge is this:
Within the simulation-first scope of this article, the declared structural extraction of a Game must be Structurally Playable.
In other words, the extraction’s Rule-constituted Possibility Space must contain somewhere two admitted Contribution Profiles that differ within a finite Role block, agree everywhere else, and open different Continuations.
This is the necessary structure I was looking for. It leaves the rest of the Game wonderfully unresolved. Nothing here says the supporting contribution is wise, satisfying, expressive, fair, difficult, informed, or even noticed. Those judgments require more relations. The one relation kept here is the one without which the Operating Instance can still run, but every admitted contribution remains ceremonial.
Why does the condition only point one way?
Because every attempt to make it sufficient has to bring something else into the room.
The usual answer is to find a form that satisfies the condition but does not feel like a game, then patch the definition until that form goes away. The patch is rarely free. It normally imports a goal, an intention, a preference, a cultural category, or an expected experience, then lets that imported relation quietly decide what the Game was supposed to mean all along.
A goal can distinguish a contribution someone prefers from a fork nobody values. An intention can explain why a Rule was authored. A culture can decide which forms it recognizes and which ones it does not. An observer can find one Trajectory profound and another unbearable. All of that may matter more than anything in this article. None becomes part of the necessary relation merely by being important.
So the bridge rules one direction out. It does not manufacture the entire other direction. A form with no Gameness Support fails the condition argued here. A form with Gameness Support has merely reached the point where the rest of the questions can begin.
So what does all that settle?
It settles where the argument must look.
Gameness is supported where Rules admit a contribution-sensitive difference in continuation. Mechanics express that relation. Presentation may disclose it, conceal it, delay it, exaggerate it, or make anyone care about it. Goals may evaluate it. None of those layers can substitute for the relation while leaving every Continuation constant.
It also explains why a branch diagram alone proves so little. Branches may come from internal nondeterminism. Two drawn branches may remain inspectably identical. A hundred prompts may all open one Continuation. The relevant question is always which admitted contribution can carry the form into which distinguishable future.
It makes several older arguments smaller too. Two mechanics express the same Rule only after the comparison says which distinctions it is discarding. Wood, silicon, paper, speech, and an engine may realize the same configured form when they preserve its operative structure. Showing a Role something the Rules previously withheld changes the disclosure relation, but disclosure alone does not create another contribution. And an input may remain perfectly ceremonial no matter how expensive its animation was.
From Game inward
Once a form has been accepted as a Game under that bridge, the same structure can be examined from the inside. The reachable States containing Agency points form its Gameness Support.
Support can be sparse or dense. It can depend on one Role or several. It can appear at the beginning, disappear for long intervals, return near an ending, or remain live almost everywhere. The necessary condition guarantees somewhere, not everywhere.
This is enough to recover the intuition that began the article. Some things called games feel…less game because much of their reachable form is structurally inert under the relation that makes their interventions more than ceremony.
The comparison still needs a declared granularity. One form may represent every animation frame as a State while another records only committed turns. One may expose hidden variables that another deliberately forgets. Counting before deciding which distinctions survive would reward bookkeeping.
After the comparison has placed both forms at one granularity, a finite reachable State space lets us call the share carrying Agency support structural gameness density. An infinite space needs a declared measure before the same language applies. A smaller share gives one precise sense in which one already accepted Game is less game than another. It remains a comparison of support distribution, not a verdict on quality and not a universal scalar into which every preference can be smuggled.
Density also says nothing about visitation. A large portion of the Possibility Space may be reachable and almost never actualized. An Operating Instance may spend nearly all its time in a small region. To speak about frequency across play requires a declared distribution over instances and visits; the Rules alone provide no canonical one.
The shape of support matters too. Density forgets how many Roles a witness needs, how latent its fork is, and whether its branches can reconverge. Those properties should remain visible instead of being combined with arbitrary weights and called objectivity.
Where purpose goes
Purpose enters in more than one place, so the word must say whose purpose it means.
A goal represented by the Rules can alter readings, admissibility, and consequences. Win conditions, loss conditions, scores, terminal predicates, and outcome orderings become part of the State when the configured form can make their differences operative.
Designer intent sits behind the form. It may explain why the Rules were chosen and motivate a later revision. It does not retroactively remove a possibility those Rules already admit.
Occupant purpose sits in front of the form. It can change which contribution a Source supplies and therefore which Trajectory becomes actual. It changes the structure itself only when the Rules represent that purpose as an operative distinction.
This is why “meaningful” cannot repair “choice” by adjective. Meaning may belong to the observer, a goal, a value system, a consequence, or a social agreement. Each can be important. Each needs its own structural location before it can change what the configured form admits.
Consequences
Consequence 1: No structurally “wrong” play remains
If a Contribution Profile is admitted at a reachable State, taking it is not outside the Rule-constituted play structure.
It may be strategically awful. It may annoy the other players. It may break an agreement that was never part of the Rules. It may expose a design mistake. None of those judgments removes the admitted contribution or its outcomes from the Possibility Space.
The Possibility Space is the reachable structure of the declared Rules. If those Rules admit the contribution, its result belongs there. Strategy, etiquette, agreements, and intent may condemn it without changing that membership.
“The designer did not expect that” does not alter the form already in operation. Intent may explain or motivate a later revision; until the Rules change, the admitted possibility remains admitted.
If stacking three Quartz lets a mage deal one million physical damage, the configured form admitted it. Calling the result an exploit does not reach backward and remove it from the Possibility Space. If it was never meant to belong there, the player did not select the wrong possibility. The Rules supplied the wrong one.
The reverse is less flattering. Tuning three Quartz contributes system depth only where the arrangement changes what the form admits, discloses, or can later distinguish. If every arrangement returns to the same Continuation and changes only a number no Rule reads, the surrounding apparatus is presenting a distinction rather than constituting one. It may still be a very satisfying evening in the equipment menu. Perhaps the menu has been doing more work than the system underneath it.
Consequence 2: A Trajectory is not its endpoint
The endpoint of a bounded Trajectory is merely its last committed State. It need not be a dead end under the Rules, correspond to a game ending, or display the credits.
A Trajectory may remain ongoing. An endpoint appears only after some interval of its operation has been bounded.
Call an interval opened and closed by its surrounding operation a Session. Call a bounded cycle recognized by the configured form a Run. One Session may contain several Runs; one Run may continue across several Sessions. Neither boundary implies terminality.
A terminal State is different from all three. It is a State from which the Rules admit no continuation. Closing a Session, completing a Run, displaying an ending, and reaching terminality are four relations that may coincide and often do not.
Two Trajectories can end at the same State without being the same Trajectory. Their raw histories remain different. Their inspectable futures agree only to the extent that the present State has genuinely absorbed every operative distinction between them. If a later reading, admission, or consequence depends on the route taken, then a trace of that route remains in the current State.
This captures the structural content of one complaint that choices did not matter. Two Contributions may open different Continuations while every immediate outcome still retains a route to one common later State under one finite bound. That was a real fork with possible Reconvergence. Whether the actual Trajectory follows one of those routes is another question.
A shared destination is weaker still than a shared State. It may name only one property of an endpoint while everything else differs. Story endings and credits receive no privileged status here. They participate when their representation changes what can be read, admitted, or continued.
Consequences also have a shelf life. Three Contributions may produce three inspectably different responses and then leave no distinction that any later Rule can consult. Agency existed at the earlier window. Its represented consequence ended there. A different ending may likewise be a genuinely different State even when nothing follows it; what it lacks is further reach, not reality at the endpoint.
That shelf life is not Latency. Latency asks how deep one must inspect before a fork first becomes distinguishable. It is not Reconvergence either, which asks whether every immediate outcome still retains some route to one common State. Measuring persistence would require comparing later counterfactual paths and declaring how their outcomes are paired. The configured form supplies no canonical number merely because the question is worth asking.
Consequence 3: Gameness is distributed
A Game needs non-empty Gameness Support. How that support is distributed remains open.
It may occupy only a small part of the reachable State space. Runtime, interface size, prompt count, map area, and the amount of authored content do not determine that distribution.
This is where ceremonial input becomes visible. A long interval may continue to accept commands while every admitted command opens the same Continuation. The Operating Instance is busy, the controller is busy, the animation department is certainly busy, and the Gameness Support is still absent from that interval.
Another interval may offer only two plain contributions and split the continuation immediately. The number of gestures tells us less than what the Rules let those gestures change.
A configuration surface may expose fifty adjustable values and carry no Agency if every admitted profile opens one Continuation. One plain contribution may change which actions remain admissible for the rest of a Run. The larger interface is not automatically the denser Game, and the longer consequence is not automatically the more enjoyable one. They are simply different structural facts.
This leaves two questions I find more useful than counting prompts or systems: where in the reachable form does Agency exist at all, and after a witness occurs, where does its resulting distinction cease to remain operative? The first has Gameness Support. The second still needs a declared persistence relation rather than a confident adjective.
Calling one Game less game than another therefore requires two prior choices: a common comparison granularity and a quantity taken from the resulting support. Structural gameness density is one such quantity. Agency arity, Latency, and Reconvergence describe other aspects. No arithmetic can choose their importance without importing a judgment from outside the structure.
Some Thought Experiments
For each case, ask which relation changed: the Rules, their configured form, an Operating Instance, a binding, or merely the observer.
- Suppose humanity goes extinct, leaving behind only a complete rulebook for chess. Does chess still exist when nobody is there to play it? If aliens find only a wooden chess set and invent a completely different way to use it, are they playing chess? If they instead find the rulebook and make their own pieces from granite, has the game changed, or only what it is played with?
- Four fish swim in a tank while sensors turn their movements into controls for Pokemon Sapphire, and eventually the game is beaten. Does Pokemon stop being a game because the inputs came from a fish? Does the fish have to know it is playing, or intend to play at all?
- The ball enters the net. No player is offside, yet the assistant referee raises the flag, the referee disallows the goal, and VAR is not consulted. At that moment, is this still the same game of football? Do the Rules give the referee the power only to judge what happened, or to make a ruling that counts even when the judgment is wrong? Does VAR change the Rules of football, or only how decisions are checked and made final?
- In a D&D session, the DM gives one player two turns to compensate for falling behind the rest of the table. Is this still the same game? Was the DM allowed to make that ruling under the Rules the table was already using, or did the table just add a house rule?
- Six children play by Rules they made together. Child A scores with a move that is allowed at the time. Before child B can repeat it, the group changes the Rule to forbid it. Does the new Rule erase A’s earlier score, or does it apply only from that moment onward? If B performs the move anyway, is it valid now? What changed between the two attempts? What game are they playing now?
On the Player Side
The same person may occupy two relations: observer and Agent. Interpretation belongs to the first; an admitted Contribution belongs to the second. Interpretation may alter a later policy, but it does not rewrite the committed past.
Holding the configured form, bindings, and Contributions fixed, swapping an observer whom the form does not represent changes neither the Possibility Space nor the Trajectory that occurred.
What a Role is shown belongs to the configured form. What its occupant has inferred but the Rules do not represent remains with the occupant. That is why choosing, guessing, knowledge, and Agency cannot be synonyms. Where the Rules conceal, a Role may expose continuation-changing alternatives while its occupant lacks any policy that targets the hidden situation. The structural difference exists; informed control begins one epistemic layer later.
Saturation
Once information has been separated from Agency, another intuition can be named without mixing the two back together.
Fix one granularity at which a Source is trying to understand the Possibility Space. The Source begins with several structural models that remain compatible with what it has received. As readings arrive and inferences are made, some models disappear. Call the Source Saturated when every remaining model is structurally the same at that chosen granularity.
Saturation collapses uncertainty about the form. It does not settle which hidden State is actual now, which goal the Source values, which Contribution is best, or whether the Source can execute it. It also leaves Agency where it was. The same Game can be fresh to one Source and Saturated for another without changing a single Rule.
The Rules may participate in that learning by disclosing evidence through their scoped interfaces. If the form represents what has been learned and lets it affect later admissions or continuations, the learning state belongs to the Game. If it does not, the update remains with the Source.
A Rule mutation already admitted by the configured form does not surprise a Saturated Source merely by occurring; its possible results were already among the structures being considered. Revising the form from outside those represented possibilities produces another configured form, and knowledge sufficient for the first need not Saturate the second.
This gives one account of familiarity without pretending that the Possibility Space shrinks whenever somebody understands it. What changes is the set of models the Source still considers possible. The configured form remains where it was, quietly waiting for somebody else to discover the obvious all over again.
Solving adds an objective, information conditions, and a solution concept. Concealment adds uncertainty about which hidden State is actual now. Knowing the whole map settles neither question by itself.
An observer-dependent feature becomes structurally relevant when readings, admissions, or continuations make it operative. Otherwise the thing under evaluation includes the observer and context in addition to the Game.
Here I take meaning to arise through the relation between Rules, presentation, context, and the observer. One player may find a Trajectory heroic, another offensive, boring, sacred, cruel, hilarious, or meaningless. Neither response rewrites the past; neither changes the configured form while it remains unrepresented there.
A designer may construct the conditions of an experience through Rules, presentation, information, rewards, punishments, and consequences. An unrepresented response is not a value the designer can commit inside the Game. There is no checkbox for mandatory interpretation, however convenient it would be.
And no, plot, messaging, and story do not clash with games. They participate structurally when the form makes them operative through readings, admissions, or continuations. Presentation may itself be disclosed by Rules; the question is what reaches which Role, not whether the material looks narrative.
Adapting a film or novel into a game is not a matter of distributing the play button across a shooting gallery, an axe-throwing segment, three yellow-painted ledges, and a giant pipe-shaped hallway hiding a loading screen. Button prompts do not establish different Continuations. Where every admitted Contribution still opens the same one, choreography remains choreography.
Just respect the intelligence and stature of the player, just as I respect you, and you respect me, alright?
Conclusion
So, after all, why do we need this? Well, it appears to make some invalid thing stay out of the way, as we need to focus on the right track.
In reality, we do not need to settle every possible definition before taking one step closer to the thing itself. We only need to stop mixing every layer together.
I think, as Indie game dev, we should keep one thing simple: before trying to make a game fun, moving, groundbreaking, artistic, addictive, and everything else, make sure the part expected to operate as a game is not structurally inert under the necessary condition argued here. Passing it is not sufficient for quality or even, by itself, sufficient for gamehood.
Recreating your imagination is not a matter of drawing a map, throwing a story and a mountain of content into it, then adding button prompts. Any configured form I am willing to call a Game here must have Rules, a Possibility Space, and non-empty Gameness Support; a Trajectory appears only when an Operating Instance actually runs it, you know, all the bullshit I have been rambling about above.
I am not trying to provide every answer. I am trying to make the question precise enough that the next answer has somewhere solid to stand. Every distinction up there is doing a job, and the only reason any of them exists is that without it a sentence I needed could not be said at all. If I ever catch one just standing around looking clever, it goes. If you want to understand the thing you are passionate about before deciding what else to do with it, I hope this gives you a perspective worth pondering.
FM39hz, 02/01/2026