TicTacToe Ring

i.e. simple tictactoe game sandbox for crash testing ai-bots designs

 

The purpose of what's above, despite fun, was to facilitate the testing of bots prototypes performance -by benchmarking them with searchable algorithms. The following have been accomplished:

• playground / tournament modes providing respectively touch'n'feel / at-scale-simulated gameplay;
• referential bots implementations i.e.: minmax, debuts, random, search (and humanxd);
• from-scratch AI-based bots prototypes (waiting for a dose of spare time...)
• transparent API exposed to interfere with: BoardAPI and PlayerAPI;
• twofold UI, via: shell CLI (see demo.gif) or clickable web GUI (once hosted here ).

 

TicTacToe is deterministic zero-sum game with 19683 possible states and 255168 gameplayes. See complete map of game tree depth depiceted by XKCD .

#QUICKTOUR:

# for CLI --script howto:
# (optional) python3.8 -m pip install tqdm==4.51.0
python3.8 -m tictactoe -o human -x minmax
python3.8 -m tictactoe -o search -x minmax -n 10  # tournament

# for GUI --server deploy & run:
python3.8 -m venv .env && source .env/bin/activate && python3.8 -m pip install -r requirements.txt
streamlit run streamgui.py

# for GUI --server deploy & run with docker:
sudo docker build -t streamgui .
sudo docker run -p 8501:8501 streamgui