If you use Monte Carlo Tree Search (MCTS) to implement an AI…
If you use Monte Carlo Tree Search (MCTS) to implement an AI agent to play a two-player, zero-sum game, you will need a playout (rollout) policy. The playout policy determines how the simulation proceeds from a newly expanded node until a terminal state is reached. It is very important that this policy is not random, since it does not model a realistic opponent and makes MCTS struggle to identify critical game-ending scenarios. For chess, one example of a non-random playout policy consists of picking moves according to the following criteria: (1) move to capture a piece, (2) move to avoid immediate capture , or (3) move randomly . Connect Four is a two-player, zero-sum game in which players take turns dropping colored pieces (each player plays with one color) into a six-row, seven-column vertically suspended grid. The pieces fall straight down, occupying the lowest available space within the chosen column. The objective of the game is to be the first to form a horizontal, vertical, or diagonal line with four pieces of the same color. The image below illustrates a game won by the player with the red pieces. Please outline a playout policy for the Connect Four game.