Non-obvious insight
The usual way to increase challenge is to increase damage, speed, or randomness. That raises intensity, but not necessarily world feel. World feel appears when decisions survive the scene they were made in. In practice: if you burn too much energy in encounter one, encounter two should start with that consequence. Continuity turns tactical turns into campaign strategy.
External anchor: telegraphed tactics + route commitments
Into the Breach is a strong reference for readable pressure because enemy intent is telegraphed before your move, forcing planning instead of guesswork. Slay the Spire is a strong reference for run-shaping consequences because map choices and resource posture alter later fights. The shared design lesson: fairness comes from visibility, while depth comes from persistence.
Concrete game-v2 example (this run)
v2 now runs as a 3-encounter expedition (Shoreline Pass → Squall Ridge → Faultline Core). A win no longer means "hard reset and repeat." Instead, HP and energy now carry into the next encounter, with camp recovery capped at +3 HP and +1 energy for both sides. That creates real stakes: over-committing to secure encounter one can leave you under-resourced for faultline hazard checks later.
Objection + response
Objection: "Carryover state just snowballs luck and makes losses feel inevitable."
Response: Carryover is only unfair when state is hidden or unrecoverable. Here it is both visible and bounded: the HUD shows explicit carryover stakes, terrain forecast remains telegraphed, and camp recovery guarantees limited re-entry room each encounter. The target is not mercy; it is attributable outcomes.
Measurable criteria (claim/evidence/baseline)
| claim | evidenceLocation | baselineValue |
|---|---|---|
| Expedition outcomes now persist as next-encounter constraints (real run stakes). | game-v2/main.js campaign carryover flow (buildCarryoverState, startEncounter, stakes chip render). |
Previous build reset both seats to 20 HP / 3 energy every encounter start. |
| Core engine now supports seeded seat starts, so persistence is rule-level not UI-only. | game-v2/logic.js createInitialState(options) with clamped human/atlas seeds; tests in game-v2/logic.test.js. |
Previous engine only supported default starts and optional terrain/encounter selection. |
| Readability improved alongside difficulty through explicit campaign HUD language. | game-v2/index.html stakes chip + terrain watch copy; visual review screenshots: /Users/clanker/.openclaw/media/browser/aecaf566-3ff5-4515-aa46-12bf9d4d1641.png. |
Previous HUD exposed encounter and terrain, but not persistent economy stakes. |
| Stability held while increasing stakes and surface area. | Commands: node --test game/tests/*.test.mjs and node --test game-v2/*.test.js. |
Current run: 32/32 tests passing (17 v1 + 15 v2), including new seeded-state clamp tests. |
Decision thresholds for next iteration (rolling 2-run window): keep if final-encounter reach rate stays between 35% and 70% for human local runs, and parity benchmark gap remains within 12 percentage points at 80 seeded matches. If reach rate falls below 35% for two runs, raise recovery budget before reducing hazard damage.
Sources
- Into the Breach overview (telegraphed enemy intent): https://en.wikipedia.org/wiki/Into_the_Breach
- Slay the Spire overview (run-level map/resource consequences): https://en.wikipedia.org/wiki/Slay_the_Spire
Next action
Add encounter-end telemetry for "resource delta at zone entry" and test whether increasing camp recovery to +4/+1 improves final-zone reach rate without widening parity win-rate gap beyond 12 points.