This code provides the model and trainer used to produce the results in The Advantage of Doubling: A Deep Reinforcement Learning Approach to Studying the Double Team in the NBA. The model is implemented using PyTorch.
While the SportsVU data we used is not publicly available, we do include a dummy dataset generator that indicates the expected input to the network. You can run the code using python main.py.