What it is. A choice is real when the options differ on an axis the player values, the stakes are legible before committing, and no option dominates the rest. Everything else is a menu with one right answer wearing a costume.

Player fantasy / why it’s fun. “This is mine.” Ownership comes from having weighed something and paid for it. The player wants to author a path, defend it, and see it play out — not to be quizzed on what you already decided was correct.

The three tests

A decision must pass all three. Fail any one and it degenerates into antipattern-fake-choice.

Distinction without legibility is a gamble. Legibility without non-dominance is a formality. Non-dominance without distinction is a coin flip. You need all three.

The core levers

Tune these to turn a fork from decorative into load-bearing.

• Opportunity cost. The pick must close a door, not just open one. Balatro’s shop makes you skip a joker to afford a reroll; Slay the Spire’s card removal costs the gold you’d spend on relics. If saying yes to A costs you nothing, A is not a choice — it’s a pickup.
• Information. How much does the player know at commit time? Full information → tactics (anchor-into-the-breach shows every enemy’s next move). Partial → read-and-bet (anchor-slay-the-spire shows the map but not the fight). None → gamble. Slide this knob deliberately; see system-telegraphs.
• Reversibility. Can the player undo, respec, or re-roll? Cheap reversal lowers the stakes and invites experimentation (anchor-into-the-breach rewind). Permanent commitment raises them and makes the choice matter (anchor-reigns can’t take back a ruling). Neither is correct — pick to fit your session length and see system-save-and-checkpoint.
• Context-dependence. The best answer should move with the situation. This is what kills dominance: anchor-into-the-breach and anchor-slay-the-spire work because the correct play depends on the board you were dealt. See system-counter-systems.

Where distinction comes from

Numbers alone rarely distinguish. Reach for a categorical axis:

If two options differ only in magnitude, merge them or re-skin one onto a new axis.

Twist seams

• A branching narrative choice, but the game never tells you which faction it fed (vector: legibility → delayed). anchor-reigns and genre-narrative-decisions live here — the tension is you feel the stakes without seeing the ledger. Powerful, but one notch too opaque and it becomes antipattern-guess-the-designer.
• A deckbuilder draft, but every card you take is also removed from a shared pool your rival drafts from (vector: non-dominance → adversarial). Opportunity cost becomes denial; the optimal pick shifts every turn because the opponent’s board is context. See system-coop-and-competition.
• A build choice, but reversible for free until the first boss, then locked (vector: reversibility → time-gated). Front-loads experimentation, back-loads commitment — the choice starts as play and hardens into identity. Pairs with system-boss-design as the lock point.

Scaling: how many, how often

More forks is not more depth. It’s the fast road to antipattern-decision-paralysis and antipattern-feature-soup.

• Trivial acts must not be forks. Opening a door, walking left — if the player can’t articulate a reason to care, don’t make them choose. Reserve decisions for moments that carry weight.
• Cluster the load-bearing ones. anchor-slay-the-spire‘s real decisions are the map node, the card reward, and the shop — three per act, each spaced. The moment-to-moment is execution, not deliberation. See pattern-pacing-and-tension.
• Cap the option count. Three to five distinct options beats twelve near-duplicates. anchor-hades shows a boon and asks you to pick from a handful, not a spreadsheet.
• Let earlier choices constrain later ones. A build that closes doors reduces future paralysis — the space narrows as identity forms. See system-progression, system-skill-trees.

Overdone / failure modes

• The illusion fork. Both paths converge to the same node. Fine as flavor, fatal as advertised agency — this is antipattern-fake-choice.
• The dominant answer. One pick is strictly best; the “choice” is a knowledge check for new players and a no-op for everyone else. antipattern-boring-optimal.
• The invisible ledger. Consequences are real but wholly unpredictable — the player picks blind and feels cheated by the result, not authored. antipattern-guess-the-designer.
• The paralysis wall. So many legible, distinct, non-dominant options that the player freezes. Depth you can’t parse is not depth. antipattern-decision-paralysis.
• The solved fork. Distinct and legible on day one, but the community finds the one line and the space collapses. antipattern-solved-metagame — design for a shifting best answer, not a fixed one.

Checklist

• Every option differs on an axis the player values, not just in magnitude.
• The player can predict each outcome’s shape before committing (or the fog is deliberate and telegraphed).
• No option dominates across all contexts — the best pick moves with the board.
• Each pick has a real opportunity cost; something is given up.
• Reversibility is chosen on purpose to fit the session, not left to accident.
• Trivial acts are not forks; decisions are clustered where weight lands.
• Option count is small enough to hold in the head.

Cross-links

Sits atop pattern-risk-reward (one axis of distinction) and feeds system-build-diversity (choices that accumulate into identity). Its shadow is antipattern-fake-choice; its overload is antipattern-decision-paralysis; its decay is antipattern-solved-metagame. Legibility leans on system-telegraphs and pattern-readability; stakes lean on system-economy. Prove the fun of a fork against design/FUN.md — a choice is real only if removing it makes the game measurably poorer.