TD-Gammon enabled a computer to outperform humans in the game of backgammon. I left out a crucial part of how TD-Gammon was trained. It did not learn through the trial and error of endless games of backgammon against a human expert. If it had done this, it would never have learned, because it would never have won. TD-Gammon was trained by playing against itself. TD-Gammon always had an evenly matched player. This is the standard strategy for training reinforcement learning systems.
