Risk / Reward

What it is. A decision has teeth when its upside is priced in a real, legible downside — push deeper for more loot at rising odds of losing it all; take the strong card that also feeds your discard pile; heal now or bank the mana. The pattern is the enemy of the no-brainer: if one option is strictly better, there was never a choice.

Player fantasy. “Do I dare?” The pull of the greedy line and the dread of the cost, held in the same breath. Owning the outcome — the win you earned by holding your nerve, the loss you earned by one more push.

Why it works

A choice is only a choice if both sides cost something. Strictly-better options are content the player clicks past; double-edged options are content the player thinks about. Depth is measured in agonised pauses.
Push-your-luck manufactures tension for free. Each “one more” raises the stake and the dread on its own — you don’t script the peak, the player’s greed builds it (see pattern-pacing-and-tension).
Variance is fun when the player chose it. A gamble you opted into reads as drama; the same swing forced on you reads as unfair. Risk must be elected.
The bank point is the whole tension. The moment you can lock in gains is where push-your-luck lives — one-more-push against cash-out-now. Remove the bank point and it’s not a gamble, it’s an outcome; the choice of when to stop is the game.

Levers

Lever	Effect	Example
Bust probability	How fast the greedy line collapses	Each extra room: +15% ambush
Reward slope	How steep the payoff for pushing	Pot doubles each step; last multiplier is the sweetener
Bank point	When you can lock in gains	Cash out at any chest; commit at the door
Double edge	The cost baked into the good option	The power card also damages you (Reigns’
Information	How much you know before committing	Telegraphed odds vs blind draw (`pattern-readability`)
Loss size	What a bust actually costs	Lose the run’s haul vs lose one heal

Applied across genres

Genre	The teeth
Narrative decisions (`genre-narrative-decisions`)	Every swipe double-edged; meters between two ditches;
Roguelike (`genre-roguelike`)	The cursed item, the risky shortcut, the elite room for elite loot
Deckbuilder (`genre-deckbuilder`)	Powerful cards with a downside; the skip-the-reward-for-tempo choice
Physics arcade (`genre-physics-arcade`)	The guarded multiplier peg; last-ball-for-the-jackpot pressure
Blackjack / push-your-luck (`anchor-balatro`)	Hit or stand; the score you have vs the score you could
Farming sim (`genre-farming-sim`)	Plant the long crop that might not ripen before the season turn (FUN.md §15)
Tactics (`genre-tactics`)	Overextend for the kill vs hold the defensible tile

Overdone when…

The no-brainer. One option dominates in every state — it’s not a decision, it’s a speed bump. Cut it or add a cost.
Fake risk. The “gamble” always pays (odds too kind, loss too small) — players learn to always push and the tension evaporates. The bust must bite.
Swing with no agency. Variance the player didn’t choose (a coin-flip boss, an unblockable spike) reads as unfair, not thrilling. FUN.md §11: an incoming spike must be blockable in principle, or it’s a coin flip, not a decision.
One choice ends the game. A single decision with an unbounded downside is a trap, not a risk. Reigns caps single-choice effects (|Δ| ≤ 20) so stewardship, not one swipe, is the game.
Risk with no signposting. A gamble whose odds and stakes the player can’t read isn’t a decision, it’s a blind draw — the tension needs the number shown (pattern-readability). Elected risk requires informed consent.

Verify / feel-gate link

Risk/reward is provable as a skill-delta with bounded stakes:

Judgement beats recklessness. The Reigns proof (FUN.md §21): a balanced policy survives, always-push loses — assert judgementScore ≫ recklessScore (19/20 vs 0/20). If reckless play competes, the risk has no teeth.
Bounded double-edge. The content lint for two-sided choices — every effect |Δ| ≤ 20, no no-op choices, every needs-flag settable (FUN.md §21). A choice that’s all upside is a lint failure.
Blockable-in-principle. For combat variance, assert the worst incoming ≤ the player’s answer ceiling (FUN.md §11) — the difference between a hard choice and a coin flip.

Worked micro-example

“A deckbuilder reward screen where the choice actually hurts.” The naive screen offers three cards, pick one — but the strong card is strictly strong, so it’s a no-brainer, not a choice. Add teeth three ways: (1) the strong card carries a double edge (it bloats your draw or damages you); (2) a skip-for-tempo option (take nothing, keep the deck lean); (3) information — telegraph the next elite so the player weighs power-now against consistency-later (pattern-readability). Now every reward is a decision. Prove it: the greedy pilot lands in the win-rate window, and a never-skip / always-grab pilot drops out of it — the cost was real (FUN.md §11).

Composes with

pattern-feedback-loops — whether to bank a lead or compound it is a push-your-luck decision riding on the loop’s gain.
system-reward-schedules — variable-ratio payoffs are risk/reward with the odds set by the drop table; keep them ethical, not exploitative.
system-build-diversity — double-edged upgrades are what make many builds viable; a strictly-best upgrade collapses the build space.
pattern-pacing-and-tension — push-your-luck is a self-building tension curve.