AI opponent for the strategic board game Chain Reaction.

$ python3 play.py --help
usage: play.py [-h] [--minimal] [--c-backend] enemy

Chain Reaction

positional arguments:
enemy        Opponent to play with - [human, random, mcts, minimax]

optional arguments:
-h, --help   show this help message and exit
--minimal    Play in a minimal non-animated window
--c-backend  Use c for processing

Here is a list of agents you can play against (in ascending levels of difficulty)

1. Random : Just a random move maker that picks from valid moves.
2. MCTS : Naive Monte Carlo Tree Search that simulates games randomly. Needs long time to be good enough.
3. Minimax : Simple DFS to minimize loss. Very hard to beat because of the aggressive static board evaluation function.

Agent Configurations can be found in config directory. The graphical configurations are stored as global variables in window.py

C backend modules are written to increase the search speed dramatically. To compile them and copy the libraries to the required folder, run ./build.sh.

• Two players take turns to place orbs of their corresponding colors. A player can only place an orb in an empty cell or a cell which already contains colored orbs of his own. When two or more orbs are placed in the same cell, they stack up.
• The critical mass of a cell is equal to the number of adjacent cells i.e., 4 for usual cells, 3 for edge cells and 2 for corner cells.
• When a cell is loaded with a number of orbs equal to its critical mass, the stack explodes. As a result of the explosion, all the orbs from the initial cell fly to adjacent cells. The explosions might result in overloading of an adjacent cell and the chain reaction of explosion continues until every cell is stable.
• When a red cell explodes and there are green cells around, the green cells are converted to red and the other rules of explosions still follow. The same rule is applicable for other colors.
• The winner is the one who eliminates other player's orbs.

A game against minimax agent looks like this -

• python3
• numpy
• pygame

• setuptools
• gcc (not tested on others)

Tested on python 3.6 and pygame 1.9.6 on both Mac and Linux

The project is completely open-source under MIT Licence, so feel free to fork and modify this code.