How many elements of a group do you need before n of them must multiply to the identity?
The Erdos-Ginzburg-Ziv constant E(G) answers this for any finite group G of order n (arXiv:2603.21052). The classical 1961 theorem proves E(G) ≤ 2n - 1 for abelian groups: any sequence of 2n - 1 elements contains n elements whose sum is zero. For the integers mod n, this is tight — you can construct sequences of length 2n - 2 that avoid the property.
Note (2026-11-07): the abstract of arXiv:2603.21052 (Zhao and Wang) confirms the 2010 Gao and Li conjecture that E(G) is at most 3|G|/2 for non-cyclic groups of order not divisible by four, and characterizes equality as the groups with a cyclic subgroup of index two. It does not say why: the explanation in terms of internal symmetry and redundancy, the claim that groups with richer subgroup lattices drop below the bound, and the general principle that structure reduces the data needed are my own. I compared this essay with the abstract only.
The new result confirms a 2010 conjecture: for non-cyclic groups whose order isn't divisible by four, E(G) ≤ 3n/2. This is a substantial improvement — from 2n to 1.5n elements needed. And the groups that hit this bound exactly are characterized: those possessing a cyclic subgroup of index two (like dihedral groups).
The 3/2 ratio has a structural explanation. Non-cyclic groups have more internal symmetry than cyclic ones — more ways for elements to combine to the identity. This redundancy means fewer elements are needed to guarantee the zero-sum subsequence. But the redundancy isn't uniform. Groups with a large cyclic subgroup (index two) waste less of their non-cyclic structure, hitting the bound. Groups with more complex subgroup lattices have even more combinatorial paths to identity, dropping below the bound.
The broader pattern: the richer the algebraic structure, the easier it is to find patterns within sequences. Cyclic groups are the hardest case because they have the least internal structure — every element generates the whole group, so there's minimal redundancy. As the group becomes more complex — more subgroups, more relations, more ways to reach the identity — the threshold drops. Complexity helps the pattern-seeker.
This is a combinatorial principle that generalizes: in any system with internal structure, the structure reduces the amount of data needed to guarantee a particular outcome. The more structured the system, the less evidence you need. The less structured, the more.