We consider the problem of finding the probability that a random triangle is obtuse, which was first raised by Lewis Caroll. Our investigation leads us to a natural correspondence between plane polygons and the Grassmann manifold of 2-planes in real n-space proposed by Allen Knutson and Jean-Claude Hausmann. This correspondence defines a natural probability measure on plane polygons. In these terms, we answer Caroll's question. We then explore the Grassmannian geometry of planar quadrilaterals, providing an answer to Sylvester's four-point problem, and describing explicitly the moduli space of unordered quadrilaterals. All of this provides a concrete introduction to a family of metrics used in shape classification and computer vision.The issue of choosing a "random triangle" is indeed problematic. I believe the difficulty is explained in large measure by the fact that there seems to be no natural group of transitive transformations acting on the set of triangles.-Stephen Portnoy A Lewis Carroll pillow problem: Probability of an obtuse triangle Statistical Science, 1994In 1895, the mathematician Charles L. Dodgson, better known by his pseudonym Lewis Carroll, published a book of 72 mathematical puzzles called "pillow problems", which he claimed to have solved while lying in bed. The pillow problems mostly concern discrete probability, but there is a single problem in continuous probability in the collection:Three points are taken at random on an infinite plane. Find the chance of their being the vertices of an obtuse-angled triangle. This is a very appealing problem and a number of authors have tackled it in the years since. After a moment's thought, it is clear that the main issue here is that the problem is ill-posed-since there is no translation-invariant probability distribution on the infinite plane, the problem must really refer to a natural probability distribution on the space of triangles. But what probability distribution on triangle space is the right one? Portnoy [23] presented several different solutions to the problem involving distributions on triangle space invariant under various groups of transformations; Edelman and Strang [9] connect the problem to random matrix theory and shape statistics; Guy [10] got the answer 3/4 for a variety of measures, and the legendary statistician David Kendall got exact answers when the vertices of the triangle were chosen at random in a convex body [16]. Interestingly,