dense, learned vector representations of teams --- can capture meaningful similarity structures directly from game outcome data. We train a neural network with team embeddings on five NBA seasons (2019-2024), analyze the learned embedding space to discover that teams cluster by playing style and con