Adversarial Search in Three Games
Playable browser implementations of the three games from Assignment 2: Quoridor, Ghosts, and UNO. Each game ships with a simple AI opponent that mirrors the algorithms suggested in the assignment notebooks.
Pick a game
Quoridor
9x9 board · 2 players · 10 walls each
Perfect-information, deterministic 2-player game. AI uses BFS shortest-path and a path-difference heuristic with shallow minimax.
Minimax Alpha-beta BFS heuristicGhosts
6x6 board · hidden ghost types · setup + play phases
Imperfect-information 2-player game. AI reasons about hidden ghost identities probabilistically, with a setup strategy and capture/exit evaluation.
Expectiminimax Hidden info Belief stateUNO (4-player)
108-card deck · Skip / Reverse / Draw2 / Wild / Wild+4
Imperfect-information multi-agent game. AI uses a rule-based / greedy heuristic that values action cards and tracks the discard pile.
Greedy policy Card counting StochasticHow this maps to the assignment
- Source: the three games come from the Standalone notebooks
Quoridor_Assignment_Standalone.ipynb,Ghosts_Assignment_Standalone.ipynb,Uno_Assignment_Standalone.ipynb. - Rules implementation: board sizes, move legality, wall blocking, hidden-ghost reveal-on-capture, UNO card matching are all reproduced from the helper classes in those notebooks.
- AI: the agents implement the strategies suggested at the bottom
of each
my_agent()docstring (path-length heuristic, expectiminimax-flavoured evaluation, greedy play with action-card valuation). - Slides: the linked presentation matches the rubric - 3pts algorithm, 2pts heuristic, 5pts mapping to class concepts per game, plus organization and quality slides.
Class concepts demonstrated
Search
BFS for shortest path (Quoridor), minimax tree expansion with depth-limited evaluation (Quoridor & Ghosts).
Adversarial & stochastic games
Two-player zero-sum reasoning (Quoridor), expectimax-style averaging over hidden ghost identities (Ghosts), chance nodes from card draws (UNO).
Heuristic design
Path-length differential, mobility, advancement-to-goal, action-card value, hand-size pressure on opponents.