What it is. Depth from few, well-chosen elements. You cut everything the [[process-pillars|pillars]] don’t require, then tune what remains until it sings. The absence carries meaning: the empty rooms, the mechanic you didn’t ship, the second button you refused.
Why it’s fun. A small verb set the player fully understands becomes a language they can speak fluently. Mastery is legible — they can see the whole game and still be surprised by it. Elegance reads as respect for the player’s time.
The core move
One question, asked of every noun on the design doc:
Does this earn its place? — does it serve a pillar, deepen an existing verb, or is it here because it seemed cool?
If the answer is “cool,” cut it. Cool is cheap; focus is expensive. The discipline is not adding less — it’s adding, testing, and then having the nerve to remove.
| Lever | What it does | Tune by |
|---|---|---|
| One-mechanic discipline | Ship a single core verb; make everything a variation on it | Count your verbs — if a new player must learn 5 before playing, you have 4 too many |
| Earn-its-place test | Trace each element to a pillar; orphans get cut | Delete it in a build; if nobody misses it, it stays deleted |
| Negative space | Let silence, emptiness, and absence do work | Where does the game breathe? A screen with nothing on it is a design choice |
| Depth over breadth | Recombine few elements instead of adding new ones | Interaction count > element count is the goal |
Why it works
- Legibility. The player holds the whole system in their head. [[pattern-readability|Readability]] is trivial when there’s little to read.
- Tunability. Fewer elements means fewer interactions to balance; you can actually reach the [[antipattern-boring-optimal|no-dominant-line]] ideal. See
process-refine-and-handoff— restraint is what makes refinement finite. - The twist lands. A [[process-the-twist|twist]] is only visible against a clean background. Clutter drowns novelty.
- Emergence gets room. [[pattern-emergence|Emergent depth]] comes from few rules interacting — not many rules coexisting. [[anchor-baba-is-you|Baba Is You]] has a handful of words and infinite puzzles.
Reference lines
- [[anchor-mini-metro|Mini Metro]] — draw a line, that’s the verb. No text, no tutorial, no menus of upgrades. The negative space is the aesthetic; the map’s emptiness makes the growing tangle legible.
- [[anchor-tetris|Tetris]] — seven pieces, one grid, gravity. Nothing has been added in forty years because nothing needs to be.
- [[anchor-into-the-breach|Into the Breach]] — perfect information, 8×8 grid, three mechs. It cut fog-of-war and army-building from the strategy genre and got a deeper game.
- [[anchor-braid|Braid]] — one time mechanic per world, never mixed. Each world is negative space around a single idea.
- [[anchor-return-of-the-obra-dinn|Return of the Obra Dinn]] — one verb (identify the dead), a two-color palette. The restraint of the whole thing is the elegance.
- [[anchor-katamari|Katamari Damacy]] — roll things up, get bigger. One rule, a whole cosmos of consequence.
Twist seams
Restraint is a lens you point at another idea — the twist is what you refuse to add.
- A deckbuilder but the deck never grows (subtraction vector) — you only ever remove and upgrade cards, never add. Scarcity becomes the whole meta. Cf.
genre-deckbuilder,system-build-diversity. - A survival game but with one resource (compression vector) — no wood/stone/food/water spaghetti; a single meter you feed. Pairs against
antipattern-currency-spaghetti; seesystem-resource-loops. - A metroidvania but with no map screen (removal vector) — navigation lives in your memory of the space, so levels must be legible by architecture alone. Cf.
genre-metroidvania,system-map-and-navigation. - A tactics game but with no hit-points (binary vector) — units live or die; every attack is lethal or nothing, so positioning is everything. See
anchor-into-the-breach,genre-tactics. - A boss rush but one button (input vector) — the entire skill ceiling lives in timing one input. See
recipe-one-button-boss-rush.
Ties to the pillars
Restraint is the enforcement arm of process-pillars. The pillars say what the game is about; restraint is the willingness to act on that by cutting what isn’t. A pillar you won’t defend by cutting is just a wish. Run the loop: [[process-intent-to-brief|brief]] → pillars → this test → [[process-composition|compose]] the survivors.
Overdone when
Restraint starves. Subtraction is a tool, not a religion — cut past the pillars and you get an empty, repetitive husk that reads as unfinished, not elegant.
| Symptom | Diagnosis | Fix |
|---|---|---|
| Player exhausts the game in 20 minutes | You cut [[system-progression|progression]] and [[system-mastery-curve|mastery]] with the fat | Add depth inside the verb, not a second verb |
| ”Minimalist” but boring | Restraint became [[antipattern-content-desert|content-desert]] | Negative space needs something to be negative around |
| Every level feels identical | One mechanic, zero variation | Recombine and escalate the one verb — see pattern-escalation-and-payoff |
| Elegant but forgettable | You cut the [[pattern-surprise-and-delight|surprise]] too | Keep one thing that breaks the pattern |
The elegant-but-empty failure is a real ceiling: [[anchor-mini-metro|Mini Metro]] survives on procedural escalation, not novelty. Without a rising [[pattern-pacing-and-tension|tension curve]], one mechanic goes stale fast.
The opposite failure
The far more common sin is the other direction: adding because you can. That’s antipattern-feature-soup — pillar-less accretion where every feature is defensible in isolation and the whole is mush. If you catch yourself justifying a feature by “players might like it” rather than “a pillar needs it,” you are already in the soup. Restraint is the antidote; antipattern-second-system is what happens when you forget it on the sequel.
Related failure modes restraint prevents: antipattern-false-depth (many shallow systems faking one deep one), antipattern-decision-paralysis (too many options, no clear read), antipattern-currency-spaghetti (resources multiplied past comprehension).
Checklist
- Can you name the game’s pillars in one breath? If not, cut until you can.
- Can a new player describe the core verb after 30 seconds? (
system-onboarding) - Have you deleted at least one feature you were fond of?
- Does every screen have room to breathe, or is it wall-to-wall?
- Is your depth coming from interactions between few elements, not the count of elements? (
pattern-emergence) - Did you keep exactly one thing that surprises? (
pattern-surprise-and-delight)
Prove the restraint held. A focused design is felt, and feel is verified downstream — check design/FUN.md for the fun-loop gates. If the game reads as elegant on paper but plays as thin, the cut went too deep; if it reads as busy, it didn’t go deep enough.