opponent-modeling

This Rock Paper Scissors AI project combines machine learning and data science techniques to create a highly effective player. It uses an LSTM neural network for move prediction, online learning for real-time adaptation, and pattern recognition. The AI employs opponent modeling, ensemble decision-making, and dynamic learning rates .

Robust opponent modeling via adversarial ensemble training — core algorithm built on evolutionary RL, plus a modified multi-particle game environment.

Opponent exploitation in imperfect game using Neural Networks.

An evolving No-Limit Hold’em AI engine for heads-up poker. It begins with rule-based decision bots, generates data for supervised learning models, and advances toward reinforcement learning agents capable of real-time, opponent-aware strategy and optimal decision making.

Automated Negotiation Agent

Chess engine that profiles an opponent's playing archetype (aggressive, materialistic, positional, tactical, passive) and adapts counter-strategy via Stockfish UCI parameters and MultiPV move selection. FastAPI backend streams simulated matchups over WebSockets to a browser dashboard.

Research chess opponent model trained on human games to reason about likely human mistakes.

Implementation of a Q-Learning agent that utilizes Opponent Modeling Algorithm

Empirical study of strategic structure and predictive policy in professional StarCraft II replays

Exact full-tree benchmarks for risk-bounded opponent adaptation in Kuhn and Leduc poker.

Heads-up poker agents by reinforcement learning (Q-learning, PPO, CFR in progress), with a conditional diffusion model that estimates the opponent hole-card distribution and feeds it back into the policy state

NegoLog V2: build negotiation agents, compare opponent models and run automated tournaments in Python. IJCAI 2024.

Argentine truco engine (1v1) with a family of AI opponents: rules → tabular Q-learning → deep RL → PIMC (Perfect Information Monte Carlo). Runs 100% locally. The finding: inference beats training.

Automated Negotiation Agent

Game AI project implementing MinMax with opponent models for Domino and Tic-Tac-Toe, including dynamic elimination of inconsistent adversary models and statistical performance analysis.

Adversarial game AI that models your playstyle and exploits your weaknesses — PyTorch opponent modeling

CBOM learns preferences from conflicting offers and uses them in human-agent negotiation. Python and Java code for the Applied Intelligence 2023 paper.

BSI-PT algorithm in the paper "Opponent Exploitation Based on Bayesian Strategy Inference and Policy Tracking"

An adaptive negotiation strategy using emotional feedback, opponent behavior and time pressure. Code for the IVA 2025 paper.

Solver combines emotional feedback, reciprocal offers and time pressure in human–robot negotiation. Code for the AAMAS 2021 paper.