A tournament is a way of orienting a complete graph: take a set of points, and for every pair of points, assign a direction to the edge between them (like recording who beat whom in a round-robin competition). A hypertournament generalizes this idea from pairs to larger groups. In a k-hypertournament, instead of ordering pairs of elements, you assign a kind of cyclic ordering to every k-element subset, where the allowed orderings are exactly those that look like "even permutations" of the elements. The question is how many such structures exist when you require them to be "ultrahomogeneous," meaning that any symmetry between two finite pieces of the structure can be extended to a symmetry of the whole structure. This is a very strong regularity condition that forces the structure to look the same everywhere.
For ordinary tournaments (k equals 2), the classification of ultrahomogeneous examples is well understood and there are only finitely many. The researcher Gregory Cherlin asked whether something similar holds for hypertournaments when k is 3 or larger. The paper answers this question, but the answer turns out to be very different from the k equals 2 case.
The main result is that for every k greater than or equal to 3, there are not just finitely many but continuum-many (meaning as many as there are real numbers) non-isomorphic countable ultrahomogeneous k-hypertournaments. This is a striking contrast with the tournament case and shows that higher-arity hypertournaments are far richer and more varied than their pairwise counterparts. The proof constructs this enormous family of examples explicitly, resolving Cherlin's question in a surprising direction.