Challenging Memory-based Deep Reinforcement Learning Agents
Open-source Guandan rules engine and heuristic AI starter kit for imperfect-information game research.
Counterfactual Regret Minimization (CFR) game theory solver computing Nash equilibrium strategies for extensive-form games and poker scenarios.
Counterfactual Regret Minimization (CFR) game theory solver computing Nash equilibrium strategies for extensive-form games and poker scenarios.
Counterfactual regret minimization of two-player zero-sum incomplete-information games in rust
Paper and ecosystem hub for residual key-structure modeling across GuanDan, DouDizhu, and Gin Rummy.
Biblioteca para Python con algoritmos de resolución de juegos de mesa (minimax, MCTS, SO-ISMCTS y MO-ISMCTS)
A Manille bot: Deep Monte Carlo blueprint + SPARTA search, with a browser GUI. Beats ISMCTS-5000 by +2.97 pts/deal.
Calibrated sequential win probability from imperfect-information game replays. HistGradientBoosting, isotonic recalibration, and a local-level state-space filter, with results reported across repeated splits.
A not-so-dumb poker game-theory lab using CFR and CFR+.
n-card Kuhn poker solved with from-scratch CFR, graded by exact best-response exploitability, and independently verified against OpenSpiel
影将棋 / Shadow Shogi — 王将を秘密に隠して指す不完全情報の対戦将棋。Next.js + Supabase
Browser Doppelkopf, canonical engine contracts, and the V3 agent design for Täglich Doko.
An open-source fog-of-war (dark) chess engine: exact belief enumeration + GT-CFR search
🇺🇾 Uruguayan Truco game engine + library written entirely in Python
Six-player imperfect-information card game with deduction baselines, imitation/PPO experiments, reproducible evaluation, and a visibility-aware replay visualizer.
A game-agnostic equilibrium AI for imperfect-information games — growing-tree search with shared information sets, predictive CFR+, GT-CFR and KLUSS. Zero dependencies.
Multi-variant rummy simulator + evaluation framework for card-game AI (heuristic / RL / CFR / LLM / GNN / hybrid). See PAPER.md for the research programme.
Quarantined legacy Doppelkopf experiments and the staging area for V3 training, evaluation, and model releases.
Duren card game for the web — four bot levels, from plain reactive play to card counting and an opponent that models your hand and bluffs, plus anonymous online rooms shared by a code. One Node process, no database, no accounts.
To associate your repository with the imperfect-information topic, visit your repo's landing page and select "manage topics."