Research map / Architectures and efficiency / RNNs and normalization
AlphaHoldem: High-Performance Artificial Intelligence for Heads-Up No-Limit Poker via End-to-End Reinforcement Learning
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 AlphaHoldem against Slumbot | AlphaHoldem · Slumbot, 100,000 hands | mbb/h: 111.56 ± 16.06 | Helped |
| Use AlphaHoldem against OpenStack | AlphaHoldem · OpenStack, 100,000 hands | mbb/h: 16.91 ± 22.34 | Helped |
| Use AlphaHoldem against human professionals | AlphaHoldem · Human professionals, 10,000 hands | mbb/h: 10.27 ± 65.13 | Mixed |
| Replace the proposed state representation with vectorized card and action features (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: 78; AlphaHoldem: 1597 | Hurt |
| Represent card and action information together with 3D tensors and a single ConvNet (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: 359; AlphaHoldem: 1597 | Hurt |
| Add historical action information to the state representation (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: 896; AlphaHoldem: 1597 | Helped |
| Use Trinal-Clip PPO instead of the original PPO loss (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: Original PPO Loss: 1257; AlphaHoldem: 1597 | Helped |
| Use Trinal-Clip PPO instead of Dual-clip PPO (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: Dual-clip PPO Loss: 1308; AlphaHoldem: 1597 | Helped |
| Use K-Best Self-Play instead of Naive Self-Play (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: Naive Self-Play: 1033; AlphaHoldem: 1597 | Helped |
| Use K-Best Self-Play instead of Best-Win Self-Play (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: Best-Win Self-Play: 1024; AlphaHoldem: 1597 | Helped |
| Use K-Best Self-Play instead of Delta-Uniform Self-Play (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: Delta-Uniform Self-Play: 931; AlphaHoldem: 1597 | Helped |
| Use K-Best Self-Play instead of PBT Self-Play (ablation) | AlphaHoldem · HUNL ablation evaluation | ELO: PBT Self-Play: 892; AlphaHoldem: 1597 | Helped |
Closest papers by meaning
- Outbidding and Outbluffing Elite Humans: Mastering Liar’s Poker via Self-Play and Reinforcement Learning (NeurIPS 2026 · similarity 0.89)
- Learning Not to Regret (AAAI 2024 · similarity 0.88)
- Combining Deep Reinforcement Learning and Search for Imperfect-Information Games (NeurIPS 2020 · similarity 0.88)
- Poker-CNN: A Pattern Learning Strategy for Making Draws and Bets in Poker Games Using Convolutional Networks (AAAI 2016 · similarity 0.87)
- Preference-CFR: Beyond Nash Equilibrium for Better Game Strategies (ICML 2025 · similarity 0.87)
- PokerBench: Training Large Language Models to Become Professional Poker Players (AAAI 2025 · 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.