Helly's theorem
/ HELL-ee /
Suppose you have a big collection of convex shapes in the plane and you want to know whether there is a single point they ALL contain — one spot lying inside every one of them. Checking all of them at once sounds hard. Helly's theorem makes a startling promise: in the plane you only need to check them three at a time. If every group of three of the shapes shares a common point, then automatically all of them share a common point. A few small tests certify a global conclusion.
The precise statement: take a finite family of convex sets in d-dimensional space. If every d + 1 of them have a point in common, then the entire family has a point in common. In the plane d = 2, so the magic number is d + 1 = 3 — every triple intersecting forces the whole collection to intersect. On a line (d = 1) the number is 2: a family of intervals all overlap as soon as every pair overlaps. The convexity hypothesis is doing essential work and cannot be dropped: for non-convex sets the theorem simply fails. The proof in the plane rests on Radon's theorem, the fact that any d + 2 points can be split into two groups whose convex hulls overlap — a tidy combinatorial seed from which Helly's result grows.
Helly's theorem launched the whole field of combinatorial convexity and shows up wherever you must coordinate many constraints — in optimization (when is a system of convex constraints jointly feasible?), in computational geometry, in transversal and piercing problems, even in voting and fair-division questions. The honest caveats are two. First, the bound d + 1 is sharp: in the plane, checking pairs is NOT enough — you can find three convex sets that pairwise overlap yet have no common point (think of three long strips arranged like a triangle's sides). Second, the basic theorem needs FINITELY many sets (or extra compactness assumptions) to be safe; with infinitely many sets you must add a closed-and-bounded condition or the conclusion can leak away.
Three vertical strips and you want a common point. Suppose strips A, B, C are placed so that A and B overlap, B and C overlap, and A and C overlap — but the three pairwise overlaps sit in different places, so no single point is in all three. This shows checking PAIRS is not enough in the plane. Helly's theorem says you must check TRIPLES (d + 1 = 3): if every triple genuinely shares a point, only then is a common point guaranteed.
Pairwise overlap is too weak in the plane; the right count is triples, exactly d + 1.
The magic number is d + 1, not 2 — and convexity is non-negotiable. Drop convexity, or check too few sets at a time, and the theorem collapses; it is a statement about convex sets specifically, not arbitrary regions.