What it is. A game whose entire depth lives in the interaction of a few rigid rules over a small readable state — Go, Chess, Hex, Onitama. No content treadmill, no hidden numbers: the rules are the game, and mastery is finding the tree they hide.

Player fantasy / why it’s fun. You saw three moves further than your opponent. The pull is legible mastery — every loss is your fault, every win is earned, and the ceiling is bottomless because the branching, not the art, is the content.

Pillars (exactly 3)

1. Few rules. The whole ruleset fits on a card. Every rule you add is a lifetime tax on every future player — spend them like they cost blood.
2. Deep interaction. Depth comes from pieces combining, not from piece count. Two rules that multiply beat ten that add. See pattern-emergence.
3. Legibility. The board is fully readable at a glance; a consequence is traceable to its cause. Perfect information means the player owns every outcome. See pattern-readability.

The loop stack

Essential systems

• system-combat-model — the piece interaction is the combat model; define capture/block/convert once, precisely.
• system-counter-systems — every strong line must have a reply, or it collapses into a solved dominant strategy.
• system-difficulty-and-dda — AI ply depth or opponent ladder scales pressure; puzzle mode authors it.
• system-onboarding — the first match must teach the rule by making you use it, not by reading it. See system-telegraphs.
• system-encounter-design — for a puzzle/campaign variant, each position is a hand-set encounter with one intended solution.

Rules live in pure, deterministic logic — a single move-application + legal-move function, seedable for any AI rollout — the way examples/sokoban/ splits pure rules from view. Study sandboxes/ for a single mechanic in isolation before you wire the board.

Content & difficulty model

• No content, only depth. A shipped abstract has one board and a rulebook; longevity is the game tree, not new levels.
• Difficulty is the opponent, not the rules. Scale search depth, add an opponent ladder, or gate authored puzzles by required look-ahead — never buff the enemy’s stats. See system-difficulty-and-dda and pattern-mastery-and-flow.
• Authored puzzle mode is the low-cost content lever: hand-set positions with a proven solution turn one ruleset into a campaign. Machine-prove each is winnable, the way the project proves every puzzle level.
• Symmetry vs asymmetry. Mirror sides (Chess, Go) are the purest test of skill; small system-faction-asymmetry (different piece sets per side, à la Onitama’s card hands) adds replay at the cost of balance work.

Signature-mechanic seeds

• Abstract strategy but the board grows as you play (structure) — you place the tile you’ll later stand on; the arena is authored by both players’ history. See mechanic-stack, pattern-emergence. Reference: Tsuro / Kingdomino spatial growth.
• Perfect-info but one hidden piece per side (constraint) — 90% of the state is open; each player conceals exactly one piece’s identity, so deduction rides on top of pure tactics. Reference: Stratego, but inverted — hide one piece instead of the whole army.
• Shared clock, simultaneous commit (input) — both players pick moves blind, then reveal and resolve at once, anchor-into-the-breach-style previewed consequence turning perfect info into a mind-game.
• One rule flips mid-game (state) — a threshold (turn count, pieces captured) rewrites one rule, so opening theory can’t carry into the endgame. See mechanic-gravity-flip.
• Wordless deduction board (constraint) — no combat at all; the win is inferring hidden placement from legible tells, anchor-return-of-the-obra-dinn logic on a grid. See recipe-detective-deduction-board.

Common pitfalls

• antipattern-solved-metagame — one line dominates and the tree collapses; the whole genre lives or dies here. Playtest for a single best opening and break it.
• antipattern-boring-optimal — the correct move is obvious and joyless; if the optimal play is never interesting, add tension, don’t add rules.
• antipattern-false-depth — many pieces, no interaction; complexity mistaken for depth. Cut pieces until each one earns its rule. See pattern-restraint-and-negative-space.
• antipattern-fake-choice — branches that all lead to the same place. If two moves never diverge in outcome, delete one.
• antipattern-rng-frustration — randomness that decides the game instead of seeding it; keep chance in setup, not in resolution. See pattern-fairness-and-trust.
• antipattern-decision-paralysis — a legal-move space so wide the player can’t read it; legibility is a pillar, guard it.

Anchors

• anchor-into-the-breach — perfect information as a design feature: every enemy telegraphs its move, so the puzzle is the board state, not hidden dice. The gold standard for legible, solvable tactics.
• anchor-mini-metro — minimalist abstract clarity: a whole system read from clean geometry, escalating pressure from a tiny ruleset. The look and restraint an abstract wants. See world-aesthetic-direction.
• Also study: Chess, Go, Hex, Onitama, Hive, Santorini — each proves depth from a card of rules.

Neighbors: genre-tactics (same tactical core, more content and theme), genre-grid-puzzle (single-player authored positions), genre-deckbuilder (abstract interaction with a content axis bolted on).

Verify

Prove the tactics land against design/FUN.md#12-·-tactics — depth that resists optimization, a readable consequence for every move, escalating pressure. The rules module is pure and deterministic, so an AI rollout is your first anti-solve check: if a shallow search finds a dominant line, the design is broken before the art. See pattern-fairness-and-trust.