Four two-dimensional convex hull algorithms are explained and experimentally compared in this paper: Graham’s algorithm [1], the first to achieve the O(n log n) optimal worst-case complexity, Jarvis’s “gift-wrapping” algorithm [2], and algorithms of Eddy [3] and of Akl and Toussaint [4], both designed to be fast on uniform distributions of points. Jarvis’s algorithm is much slower than the other three; it is, however, the only one that generalizes to higher dimensions. The Eddy and Akl/Toussaint algorithms are, as expected, the best for uniform distributions. But the most interesting result of this study is that the Graham algorithm is the best overall; it is only slightly outperformed on uniform distributions, and is the clear winner when a large fraction of the given points are on the hull.