Using a reward function or reinforcement signal R(x,z) we can train the agent to play a shot along the lines or the corners of the table. A -ve reinforcement is received if the shot does not land on the table.