How soniquentin/abalone_ai_minimax works
Project type and purpose
• This is an AI project for the board game Abalone. It’s neither a web service nor a distributed system but a standalone Python application with:
• A simple GUI (HTML/CSS/JS)
• A game engine (board, game state, match manager)
• An AI agent implementing Minimax with a transposition table and heuristic functions
• Test scripts to tune and benchmark the agent
File structure
• src/GUI: static UI files (index.html, main.css, main.js)
• src/main_abalone.py: entry point that bootstraps the GUI + game engine
• src/master_abalone.py: orchestrates the match loop (Controller)
• src/game_state_abalone.py, src/board_abalone.py, src/player_abalone.py: core model classes provided by the course
• src/my_player.py: your AI agent implementing Minimax + optimizations
• src/TranspositionTable.py: caching layer to avoid re-evaluating states
• src/coords.py: coordinate-conversion utilities used by both UI (to render stones) and the agent
• tests/: scripts and data for heuristic tuning and tournament benchmarking
README for architecture cues
• The system is structured roughly as MVC:
– View: GUI (served locally)
– Controller: main_abalone.py + master_abalone.py
– Model: board_abalone.py, game_state_abalone.py
• The AI agent (“my_player.py”) plugs into the Controller as a Player subclass
• A Transposition Table caches Minimax evaluations
Essential elements to capture
• Data flow: user inputs, method calls, cache hits/misses
• Design patterns (e.g., “Minimax search loop”, “Strategy: Player subclass”)