CR_MCTS_expand_and_evaluate.py

def expand_and_evaluate(self, parent, action, child):
    if child.win is None:
        # it's either a non-terminal node or a terminal node that hasn't been
        # visited before

        # using controller (which knows the rules of the game) to get the next state
        next_state_obj, win = self.controller.get_next_state(parent.state, action)
        next_state = next_state_obj.get_array_view()
        if win is None:
            # not a terminal node
            child.state = next_state
            child.set_action_validity()
            child.initialize_edges()
            val = child.get_value()
        else:
            # it's a terminal node
            child.win = win
            # say win is 'red' and parent's turn was also red, that
            # means parent has won, which implies the child has lost, i.e., value = -1
            # for the child
            val = -1 if parent.state['player_turn'] == win else 1 
    else:
        # it's a terminal node
        val = -1 if parent.state['player_turn'] == child.win else 1

    # updating the statistics for the newly expanded node
    child.W += val
    child.N += 1

    '''
    value for the current player positively correlates with the winning chances of the current player. 
    If the current player has value +1, it means it would likely win, 
    which implies its children (and its parent, if it is not the root node), which are actually
    the opposite player, would have value -1.
    hence -1 * val is passed back
    '''

    return -val