Research map / Learning theory and algorithms / Submodular and game algorithms
Safe and Nested Subgame Solving for Imperfect-Information Games
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use nested subgame solving instead of action translation | off-tree responses in NLFH · Nested solving NLFH off-tree bet | exploitability (mbb/h): 1,465 -> 119.1 | Helped |
| add reach-aware gifts to Maxmargin solving | Maxmargin · Small Flop Hold'em, 200 buckets | exploitability (mbb/h): 30.05 -> 29.88 | Helped |
| add reach-aware gifts to Maxmargin solving | Maxmargin · Turn Hold'em, 200 buckets | exploitability (mbb/h): 427.6 -> 424.4 | Helped |
| use estimated alternative payoffs rather than blueprint best-response values | Resolve · Small Flop Hold'em, 200 buckets | exploitability (mbb/h): 60.16 -> 11.66 | Helped |
| add distributional alternative payoffs to estimated payoffs | Estimate · Large Flop Hold'em, 200 buckets | exploitability (mbb/h): 52.62 -> 49.56 | Helped |
| combine reach-aware gifts, estimated payoffs, and distributional payoffs | subgame solving · Turn Hold'em, 200 buckets | exploitability (mbb/h): 684.6 -> 116.8 | Helped |
| use gifts without splitting them across subgames (ablation) | Reach-Maxmargin · Turn Hold'em, 200 buckets | exploitability (mbb/h): 424.4 -> 333.4 | Helped |
| use gifts without splitting them across subgames (ablation) | Reach-Estimate + Distributional · Small Flop Hold'em, 200 buckets | exploitability (mbb/h): 10.21 -> 9.560 | Helped |
| scale gifts beyond the theoretically safe allocation (ablation) | Reach-Maxmargin · Flop Texas Hold'em | exploitability | Mixed |
| use Unsafe subgame solving | blueprint strategy · Large Flop Hold'em, 30,000 buckets | exploitability (mbb/h): 0.5514 -> 396.8 | Hurt |
| use Reach-Maxmargin rather than Maxmargin for nested solving | off-tree responses in NLFH · Nested solving NLFH off-tree bet | exploitability (mbb/h): 122.0 -> 119.1 | Helped |
| use estimates instead of safe blueprint values | subgame solving · NLFH and NLTH experiments | exploitability | Helped |
Closest papers by meaning
- Subgame solving without common knowledge (NeurIPS 2021 · similarity 0.93)
- Refining Subgames in Large Imperfect Information Games (AAAI 2016 · similarity 0.92)
- History Filtering in Imperfect Information Games: Algorithms and Complexity (NeurIPS 2023 · similarity 0.91)
- Safe and Robust Subgame Exploitation in Imperfect Information Games (ICML 2024 · similarity 0.89)
- Opponent-Model Search in Games with Incomplete Information (AAAI 2024 · similarity 0.87)
- The Update-Equivalence Framework for Decision-Time Planning (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.