When a game is lost, or when loss is unavoidable, we identify the critical position and the losing move made in the critical position which led to the loss. The critical position and its losing move are then added to the failure database. There are two types of failure in tic-tac-toe. First, if a move is played, and then the opponent plays a move which immediately wins, the move which allowed the opponent to win is identified as a failure and added to the failure database along with the critical position. Second, if the current player has no moves available which don't lead to failure, then the previous move by the same player is identified as a failure and added to the failure database along with the critical position.