The article introduces TD-Gammon, a neural network program designed to play backgammon by self-playing and learning from the outcomes. TD-Gammon uses Temporal Difference (TD) learning, a method that updates the network's weights based on the difference between successive predictions, to learn complex nonlinear functions. The program was developed to explore new ideas in reinforcement learning, particularly the temporal credit assignment problem, which is challenging in real-world applications. TD-Gammon's performance surpasses previous computer programs, achieving a strong intermediate level of play and even surpassing world-class human players in some cases. The article discusses the complexity of backgammon, the challenges of training a computer to play it, and the unique features of TD learning, such as the ability to learn from raw board information and the discovery of new strategies. The results highlight the potential of TD learning in both game-playing and other fields, such as robotics and financial trading.The article introduces TD-Gammon, a neural network program designed to play backgammon by self-playing and learning from the outcomes. TD-Gammon uses Temporal Difference (TD) learning, a method that updates the network's weights based on the difference between successive predictions, to learn complex nonlinear functions. The program was developed to explore new ideas in reinforcement learning, particularly the temporal credit assignment problem, which is challenging in real-world applications. TD-Gammon's performance surpasses previous computer programs, achieving a strong intermediate level of play and even surpassing world-class human players in some cases. The article discusses the complexity of backgammon, the challenges of training a computer to play it, and the unique features of TD learning, such as the ability to learn from raw board information and the discovery of new strategies. The results highlight the potential of TD learning in both game-playing and other fields, such as robotics and financial trading.