{"data":{"kind":"file","path":"README.md","version_id":"tq3vb8stx682k9hfm0lx8upm","entry":{"name":"README.md","path":"README.md","is_directory":false,"size":1339,"modified_at":"2026-08-20T13:34:23.402000","content_hash":"aee6fea0a2179b264e6345e2fc7a83ae0d958c689458920022c93fe1907aeaea"},"entries":[],"content":"# Dominion Deck-Building Environment\n\nA strategy evaluation environment modeled after the deck-building card game **Dominion**.\nTests an LLM's ability to analyze kingdom card sets, plan optimal buy sequences,\nevaluate card synergies, and project victory point outcomes under uncertainty.\n\n## Task Types\n\n1. **Buy Phase Optimization** — Given a kingdom supply and current deck state, recommend the optimal card(s) to buy.\n2. **Synergy Analysis** — Identify which cards in a kingdom set combo well together and why.\n3. **VP Projection** — Estimate final victory point totals based on deck composition and remaining turns.\n4. **Opening Strategy** — Choose the best 2-4 cost opening pair given a specific kingdom.\n5. **Counter-Strategy** — Adapt buying strategy against a known opponent plan.\n\n## Scoring\n\n- **Optimal Buy Accuracy** — Does the model pick the mathematically strongest buy?\n- **Synergy Recognition** — Does it identify real card interactions (e.g., Village + Smithy)?\n- **Reasoning Quality** — Are explanations grounded in deck-building theory (deck thinning, terminal actions, payload)?\n- **Edge Case Handling** — Does it account for pile-outs, three-pile endings, and Duchy dancing?\n\n## Usage\n\n```python\nimport verifiers as vf\nenv = vf.load_environment(\"polo/dominion-deck-building\")\n```\n\n## Version\n\nv0.1.0\n","encoding":"utf-8","truncated":false,"total_bytes":1339},"status":null}