Auto-Battler

What it is. In a prep phase you buy units from a shared pool, arrange them, and stack synergies; then combat resolves without your input and you watch your build prove itself. Win or lose the round, adjust, and buy again. Extension beyond FUN.md’s 21 genres.

Player fantasy. “I built the machine; now I watch it work.” The pleasure of composition — of finding the synergy that snowballs — plus the tension of a fight you set up but can no longer touch.

Pillars

Prep is the game; combat is the reveal. All agency lives in the shop and the board. The auto-resolved fight is feedback on a decision, choreographed, not a thing you play. (This is law 6 taken literally: the sim resolves, the view replays.)
Economy is the spine. Gold, interest, roll-vs-save, level-vs-power — the push-your-luck of the shop is where skill lives. Faucets and sinks must breathe.
Synergy over stats. Power comes from combinations (traits, tiers, positions), so many builds stay viable and no single unit is a solve.

The loop stack

Scale	The beat
Moment	Buy / sell / reroll / reposition one unit; a shop decision.
Encounter	One round: prep, then the auto-fight resolves and scores the build.
Session	A match: survive rounds vs a ladder/opponents until one economy wins.
Meta	Unlocked rosters/traits/heroes; MMR; the meta-of-comps players learn.

Essential systems

System	Why it’s load-bearing
`system-economy`	Gold, interest, and the roll-vs-save tension are the primary decision layer.
`system-build-diversity`	Many viable comps is the replay engine; synergy makes builds, not stats.
`system-unit-rosters`	A legible roster of roles/tiers is what players draft and read.
`system-counter-systems`	Comp-vs-comp near-hard counters keep the meta from converging on one build.
`system-reward-schedules`	The shared-pool draw and reroll are variable-ratio; tune the odds.

Content & difficulty model

The shared pool is a constraint, not a faucet. Contested units mean your comp affects opponents’ options; model the pool as finite (draining it is a real play).
Balance is a win-rate window over comps, borrowing the deckbuilder instrument: no comp should win too often or never — assert a band, and check both edges break.
Economy solvency is a law-3 inequality. Interest breakpoints, level costs, and round income must let a disciplined build power-spike on schedule; state and assert the payback curve so no tier strangles the mid-game.
Positioning must matter provably. The same units in a bad formation must lose to the same units well-placed — that’s the proof the board is a real decision.

Signature-mechanic seeds

Autochess but you place units on a hero’s looping path and combat auto-runs as they walk it — the board becomes a track (structure; the Loop Hero bend, composes anchor-loop-hero).
Auto-battler but the shared pool is literally shared — one deck, drafted competitively (constraint; sharpens system-counter-systems denial).
Auto-battler but you can interrupt the fight once, at a cost — a single spent moment of agency in a watch-only genre (mechanic-swap; pattern-risk-reward).
Auto-battler but co-op: two players share one economy and one board (structure; composes genre-coop-chaos, system-coop-and-competition).
Auto-battler but units level by surviving rounds, not by gold — the board is a persistent roster you nurture (mechanic-swap; composes system-progression).

Common pitfalls

Dominant comp. If one synergy wins regardless, build diversity is a lie; the win-rate window catches it — widen counters, not nerf-hammer a single unit.
Economy with no tension. If saving always beats rolling (or vice versa), the core decision evaporates. Interest and power-spikes must trade off honestly.
Combat you must watch but can’t read. If the auto-fight is illegible, players can’t learn from the reveal. Choreograph it so the reason a build won is visible.
Non-deterministic resolution. The fight must be a pure function of board + world.rng seed; then goldens, clone-and-score comp bots, and replays are free (FUN.md law 7). No Math.random in the fight.

Anchors

anchor-loop-hero — composing placement + auto-combat + deck into one novel loop; the reference for “arrange, then watch it resolve.”

Verify — extension note

Auto-battler is an extension; with no dedicated FUN.md section it composes its parents’ verify patterns:

Auto-resolved combat honesty → FUN.md §12 — Tactics: the fight is a pure sim returning choreography; 1-ply/clone-and-score comp bots are real baselines; golden end-state per board. This is the primary proof.
Draft/economy balance → FUN.md §11 — Deckbuilder: win-rate window over comps; a “never-synergise” pilot loses well below it.
Pacing solvency → FUN.md §14 — Incremental: assert flat-ish payback curves across economy tiers; no power desert.
Determinism and cosmetic-view (laws 6–7) apply unchanged — the watch phase is the clearest case of “sim resolves, view replays.”

Composes with

system-economy — the shop is the game; borrow its faucet/sink and interest tuning.
system-build-diversity — synergy space is the replay value.
system-counter-systems — comp-vs-comp counters keep the meta honest.