Monte-Carlo balance harness & CI assertions

[SorraTheOrc]

Summary: Run monte-carlo balance sweep with expanded pool and add strict, machine-checkable balance assertions in CI.

Acceptance Criteria

• Monte Carlo harness runs with expanded pool and writes aggregate metrics.
• CI enforces: no deck starvation; 100% game completion across seeds; average coin-per-turn change vs baseline within ±15%; aggregate win-rate delta within ±10%.
• If thresholds exceeded, CI fails and surface actionable metrics in results artifacts.

Minimal implementation:

• Update market.integration.test.ts or add new test to run monte-carlo sweep and assert thresholds.
• Run smaller local sweep (e.g., 200 seeds) to tune thresholds before full CI.

Deliverables: updated monte-carlo tests, result artifacts, CI gating config.

Dependencies: CG-0MP1QZF7V001TWRY (Integration), CG-0MP1QZHPO003OVU5 (Unit tests & schema).},

[SorraTheOrc]

pi

Implemented Monte Carlo guardrails and baseline drift assertions for expanded pool.

Changes:

• example-games/main-street/MainStreetMonteCarlo.ts
  • Added per run and aggregate metric
  • Included in CSV output
• scripts/generate-main-street-monte-baseline.ts
  • Generates baseline file used by CI drift checks
• docs/main-street/monte-carlo-baseline.json
  • Baseline winRate + averageCoinsPerTurn for deterministic seed set
• tests/main-street/monte-carlo-guardrails.test.ts
  • Enforces no max-turn starvation, win-rate delta <= 10%, coin/turn delta <= 15%

Validation run:

• npx vitest run --project unit tests/main-street/monte-carlo-guardrails.test.ts tests/main-street/monte-carlo-balance.test.ts tests/main-street/market.integration.test.ts

Commit: f9dae55

[SorraTheOrc]

pi

Correction to prior comment: field names are finalCoins and averageCoinsPerTurn (shell stripped inline code formatting in previous command).

Commit: f9dae55