Mastery Curve — Easy to Learn, Hard to Master

What it is. The distance between the floor (the skill to play at all) and the ceiling (the skill to play well), and how densely that distance is packed with learnable technique. “Easy to learn, hard to master” is not a slogan — it’s a measurable shape you build on purpose.

Player fantasy / why it’s fun. “I keep getting better and the game keeps having more to give.” Where system-progression grows the character, mastery grows the player. The pull is competence itself — the moment a maneuver that was impossible becomes automatic.

When to use / when NOT

Use it when	Skip it when
The game will be replayed for skill, not content	It’s a one-shot narrative or a solved puzzle set
Execution or decision depth is the core toy	The verbs are trivial and meant to stay so (some idlers)
You want longevity without infinite content	You’d rather grow the character than the player

Mastery and character-progression are not opposites — layer them. But name which one carries longevity. Tetris has zero character progression and infinite mastery; a looter has huge progression and a modest ceiling.

Variants

Variant	The depth is in	Ceiling raised by	Example
Execution	Motor skill, timing, precision	Frame-tight tech, chaining	Celeste dashes; fighting-game links
Decision	Reading state, planning	Deeper lookahead, risk math	Into the Breach; chess
Optimization	Squeezing the system	Route/build/econ mastery	Factorio; speedruns; Balatro synergy
Expression	Style within freedom	Personal flair, combos	Character-action; Peggle bank shots
Knowledge	What you know	Learned facts, not reflexes	Outer Wilds; deduction games

The best games stack two — Balatro is decision + optimization; Celeste is execution with a knowledge layer of hidden routes.

Tuning levers

Lever	What it controls	Healthy range / rule
Floor height	How hard it is to start	Low — first success in the first minute. Grace (`system-grace`) lowers the floor without lowering the ceiling
Ceiling height	How much room to grow	Tall enough that experts still find gains after hours
Depth density	Technique per hour of play	Steady discovery — a new “oh, I can do that” every session
Floor↔ceiling coupling	Do assists lower the ceiling?	Decouple. Assist mode must not cap top-end play (`anchor-celeste`)
Legibility of depth	Can players see the next tier?	Show it exists (a leaderboard, a rank, a visible-but-hard route)
Punish vs reward gap	How the game rewards skill	Skill should earn, not just avoid loss — style meters, score multipliers

The flow channel is the design target: challenge tracks rising skill so the player sits between boredom and panic. That’s a whole pattern — pattern-mastery-and-flow — this system is its concrete mechanisms.

How it wires to Hayao

Mastery is provable as a skill delta. The core proof of a real ceiling is [[FUN.md law 2]]: run the intended (skilled) strategy AND a null (unskilled) one and assert the gap — drafting 17/20 vs 9/20, braking 26.2s vs flooring 27.7s. A game with a real ceiling shows a large, reliable delta. No delta = no ceiling; skill doesn’t matter.
Pure state makes skilled bots cheap. Because world.state is plain JSON and rng flows through world.rng, you can write a skilled pilot (lookahead, aim search) and a dumb one and score both — the ceiling becomes a number ([[FUN.md law 7]]).
Grace lowers the floor for free. Coyote time, input buffering, mercy — the system-grace mechanisms — widen the entry without touching the top end. Wire them per [[FUN.md law 5]] (grace is a system, unit-tested edge-in/edge-out).
The ceiling is content you generate, not gate. Depth comes from the rules interacting (pattern-emergence), not from more levels — but a well-ramped campaign (src/content/, assertRamp) gives the floor→ceiling slope room.

Fails when…

Floor too high. Players bounce before they find the fun. Lower it with grace and onboarding, not by removing depth.
Ceiling too low. Mastered in an hour; nothing left to chase. The skill-delta goes flat — expert and novice score alike.
Fake depth. Complexity that isn’t mastery — randomness, unreadable state, input lag. Depth must be learnable, i.e. the delta must be earnable.
Assists cap the ceiling. If easy mode also caps top play, experts feel robbed. Decouple (Celeste’s assist mode changes the floor, not the ceiling).
Invisible depth. Technique nobody knows exists. Signpost that there’s more.

Verify

The skill-delta proof — intended vs null, asserted gap: [[FUN.md law 2]].
Skilled vs dumb pilots via clone-and-score: [[FUN.md law 7]].
Grace windows edge-tested (the floor’s mechanisms): [[FUN.md law 5]].
Flow-channel framing: pattern-mastery-and-flow.

Composes with

pattern-mastery-and-flow — the flow-channel pattern this system implements.
system-progression — character growth under player growth; name which leads.
system-grace — lowers the floor without lowering the ceiling.
system-onboarding — teaches the floor; hints at the ceiling.
system-difficulty-and-dda — keeps challenge inside the flow channel over time.
system-build-diversity — optimization/expression depth as many viable paths.