Physics Arcade

What it is. You aim one launcher, release, and a ball obeys clean, predictable physics — bouncing off pegs, bricks, bumpers, walls — while you watch it pay off. Input is trivial (an angle, a power); the depth is in reading the table and the delight is in the cascade you set in motion.

Player fantasy. “I called that shot.” The satisfaction of a bank-shot you planned plus the slot-machine thrill of a ricochet you didn’t — cause you own, consequence the physics authors.

Pillars

Trustworthy flight. The ball never tunnels, never teleports, never cheats. Swept collision (closed-form time-of-impact) makes the table readable — you can plan a bank shot because the sim will honour it.
Big reward on tiny input. One angle, maximal spectacle. The juice budget is the game; a plain input earns a disproportionate light-and-sound payoff.
Legible table. Every peg, bumper, and multiplier tells you what it does before you fire. The board is a puzzle you solve with an aim.

The loop stack

Scale	The beat
Moment	Nudge the aim; the trajectory preview updates; commit.
Encounter	One launch → a chain of bounces → clears + score, resolved and choreographed.
Session	A board (all targets) or a run of boards; ball budget as the fail clock.
Meta	Unlocked launchers/balls/perks; high-score chase; new table layouts.

Essential systems

System	Why it’s load-bearing
`system-reward-schedules`	The variable-ratio ricochet is the genre’s dopamine engine; tune the payoff curve.
`pattern-juice-choreography`	Impact feedback (hitstop, particles, pitch-rising combo) is the product.
`pattern-risk-reward`	The greedy multiplier peg guarded behind a hard bank; last-ball pressure.
`system-build-diversity`	Optional: launchers/balls/perks that change how the table plays.
`system-progression`	Board-to-board unlocks; score thresholds that gate the next table.

Content & difficulty model

Layout is the level. Difficulty is peg geometry, not ball speed — a tight cluster behind a wall, a lone multiplier in a corner. Author boards, not stats.
Ball budget is the fail clock. Fewer balls, tighter clears. Grace = an extra ball earned by a bucket catch or a full-clear bonus.
Prove winnability with the aim-search bot. Clone the world, fire candidate aims, count clears. This is both your “is the board beatable?” gate and a difficulty meter — a board only one aim clears is a spike.
Energy honesty is a fairness rail. A bounce never leaves the ball faster than it arrived; a conservation check catches physics bugs that masquerade as liveliness (see FUN.md §19).
Ramp by geometry, then by budget. Early boards teach one bank; later boards chain them and starve the ball count. The aim-search bot’s clear-count per board is your ramp read-out — plot it, look for cliffs, smooth them.
Skill delta is measurable. The planned-aim bot must clear boards a random-aim bot can’t (FUN.md law 2). If firing blind competes with aiming, the table has no puzzle — the depth is fake and the juice is doing all the work.

Signature-mechanic seeds

Peggle but every peg you leave alive fires back next turn — the table is an enemy that reloads (constraint + mechanic-swap; composes genre-tower-defense).
Breakout but the paddle is gravity — you tilt the whole table instead of moving a bar (perspective; the input becomes a field, not a position).
Pinball but the flippers cost mana you bank from clears — economy on the twitchiest genre (mechanic-swap; composes system-economy).
Peggle but co-op, one aims and one nudges the table mid-flight (structure; composes system-coop-and-competition, genre-coop-chaos).
Breakout but the bricks are a deck you drafted — layout as loadout (structure; composes genre-deckbuilder).

Common pitfalls

Substep-and-pray collision. Discrete steps tunnel the ball through thin walls at speed; the shot the player trusted fails invisibly. Use swept TOI — the table stops being trustworthy the first time a bank shot lies.
Difficulty via ball speed. Faster balls read as unfair, not harder. Make it harder with geometry and budget.
Juice with no sim consequence. Screen-shake on every peg dilutes the payoff; reserve the big choreography for multipliers and clears (the 2-senses contract).
RNG that isn’t yours. Deflection randomness must run through world.rng or goldens and shot-planning bots break (FUN.md law 7).
No breather. A wall of trick-shot boards with no easy clear exhausts; pace a gimme board after a spike so the reward loop can reset (see pattern-pacing-and-tension).

Anchors

anchor-peggle — variable-ratio reward + maximal juice on a trivial input; the canonical “big payoff, tiny action” loop.
anchor-tetris — the discipline of one perfectly-tuned verb; borrow its restraint, not its board.
anchor-katamari — trustworthy contact physics escalating a single verb into spectacle; steal its exponential payoff curve for multiplier chains.

Verify

Prove it against FUN.md §19 — Physics arcade: swept-collision no-tunnel proof (24k px/s + 1px-graze-misses), energy invariant, aim-search bot clears within the ball budget, golden run. See sandboxes/physics-lab for the collision primitive in isolation.

Composes with

genre-tower-defense — coverage geometry and “counter by placement” transfer when pegs fight back.
pattern-juice-choreography — this genre lives or dies on impact feel.
system-reward-schedules — the ricochet is a variable-ratio drop; tune it ethically.