How can I train a neural network with previous matches?

Question

Suppose I wanted to teach a neural network to play chess. I realized the method of learning by reinforcement. I have seen cases where these networks learn from examples of matches, but how can this be done?