After a brief introduction to the hyperbolic plane, we turn to studying the traveling salesman problem (TSP). The problem has a PTAS, and we show that a Euclidean algorithm from the early nineties solves it exactly in n^{O(\sqrt{n})} time, which is almost optimal under ETH. We then argue that the interesting instances of TSP in the hyperbolic plane should be sparse enough so that the hyperbolic geometry has an effect. A well-spaced instance is one where the pairwise distance between points is at least some constant. Using a novel separator theorem, we show that hyperbolic TSP on well-spaced point sets can be solved in n^{O(log^2 n)} time, so somewhat surprisingly, the problem may not be NP-hard.