Narrative Decisions

What it is. You steward something fragile — a kingdom, a ship, a life — through a stream of decisions, each with two edges. Every choice moves competing meters; let any meter hit a ditch and you lose. The content is the game.

Player fantasy. “I’m holding it together by judgement alone.” The weary, compulsive pull of one-more-decision, of keeping every plate spinning when every plate-spin tips another.

Pillars

Every choice is double-edged. No free wins. A decision that helps one meter must cost another — the tension is the point. A no-op choice is a dead choice.
Meters live between two ditches. Each meter loses at both extremes (too little AND too much). You’re threading, not maximising. Judgement, not a fixed policy, must win (19/20 vs 0/20 in the campaign).
Content is the level. Difficulty, pacing, and voice all live in the card deck. When content is data, editorial judgement becomes CI (see below).

The loop stack

Scale	The beat
Moment	Read a card, weigh two edges, commit; watch the meters twitch.
Encounter	A chain of related cards — an arc that pays off or backfires.
Session	A reign/run: survive until a doom fires or you reach an ending.
Meta	Unlocked arcs/characters/endings; flags carried between runs; the map of what-ifs.

Essential systems

System	Why it’s load-bearing
`world-narrative-delivery`	Story is told through the choices and meters — embedded, not narrated at you.
`system-economy`	Meters are currencies with faucets and sinks; the whole game is balancing them.
`system-emergent-systems`	Flags/relationships that remember past choices turn a deck into a personal history.
`pattern-risk-reward`	Every double-edged card is a push-your-luck bet on which meter can afford the hit.
`pattern-feedback-loops`	Meters that feed each other (low treasury → unrest → worse options) create the death spiral.

Content & difficulty model

Editorial judgement is CI. When choices are data, lint them: unique ids, bounded effects (|Δ| ≤ 20), every needs-flag actually settable somewhere, no no-op choices, no unreachable cards. The content lint is the balance test (FUN.md §21).
Doom attribution matters. Each ending must fire from ITS meter at ITS edge — a swapped ending string wrecks the fiction invisibly. Assert every doom fires its own ending, at its own edge.
Arcs terminate both ways. A multi-card arc must be able to resolve well and badly; a one-way arc is a cutscene.
Prove judgement beats policy. A balanced-policy bot survives; an always-swipe-one-way bot loses (0/20). If a fixed policy wins, the meters aren’t really ditched on both sides.
Pacing lives in the deck order. A calm-cards-only stretch is a desert; a crisis-cards-only stretch is a death march. Weight the draw so tension breathes — a scare, then a reprieve, then the arc’s payoff (see pattern-pacing-and-tension).
Every card is a small risk-reward bet. The player is betting which meter can absorb the hit this turn; the fun is that the safe-looking edge is often the slow poison. Author both edges to bite differently, never symmetrically.

Signature-mechanic seeds

Reigns but the advisors lie, and trust is a hidden meter — you’re weighing the source, not just the choice (mechanic-swap; composes system-emergent-systems).
Reigns but one deck, played by two rulers taking turns — competitive or co-op stewardship (structure; composes genre-coop-chaos, system-coop-and-competition).
Reigns but your past selves’ choices become cards for your heir — the reign is a run in a roguelite dynasty (structure; composes system-meta-progression).
Reigns but the meters are people, and each ditch is a character leaving (theme/tonal — reframes numbers as relationships).
Reigns but time-pressured — a card auto-resolves badly if you stall (constraint; adds a real-time clock over the turn-based deck).

Common pitfalls

Dominant strategy. If “always favour treasury” survives, a meter isn’t ditched on both sides — the balanced-policy proof is passing while the game is broken. Fix the ditches, not the bot.
Unbounded effects. A card that swings a meter 60 points makes planning impossible and the arc a coin flip. Cap |Δ| and assert it.
Ending soup. Dooms that fire from the wrong meter, or generic “you lost” strings, gut the fiction. Attribute each ending to its own edge.
Content that isn’t data. Hand-coded branches can’t be linted and rot silently. Keep choices in a pure data deck (FUN.md law 7) so editorial checks run in CI.

Anchors

anchor-reigns — swipe-choice narrative; meters between two ditches; stewardship as the whole verb.

Verify

Prove it against FUN.md §21 — Narrative decisions: balanced-policy bot survives; always-left loses 0/20; content lint (unique ids, |Δ| ≤ 20, settable flags, no no-ops); every doom fires its own ending; arcs terminate both ways. Chrome/menus via showScreen(); the card view is cosmetic.

Composes with

world-narrative-delivery — the story is delivered through choices; this genre is its purest engine.
system-economy — treat meters as currencies and the balance problem becomes a faucet/sink problem.
pattern-feedback-loops — inter-meter loops author the death spiral and the comeback.