A self playing Tic-Tac-Toe Game that never loses. The playing agent will look at all the possible moves and then, predict and play the best one. Upon sufficient training , even in the worst case, the match could only be ended in a draw.
Design of a learning system, Machine Learning, Tom Mitchell, McGraw Hill, 1997. https://www.cin.ufpe.br/~cavmj/Machine%20-%20Learning%20-%20Tom%20Mitchell.pdf