Hill’s Conjecture states that the crossing number cr(Kn) of the
complete graph Kn in the plane (equivalently, the sphere) is
$\frac{1}{4}\lfloor\frac{n}{2}\rfloor\lfloor\frac{n-1}{2}\rfloor\lfloor\frac{n-2}{2}\rfloor\lfloor\frac{n-3}{2}\rfloor=n^4/64
- O(n^3)$. Moon proved that the expected number of crossings in a spherical
drawing in which the points are randomly distributed and joined by geodesics is
precisely n4/64+O(n3), thus matching asymptotically the conjectured value
of cr(Kn). Let crP(G) denote the crossing number of a graph
G in the projective plane. Recently, Elkies proved that the expected number
of crossings in a naturally defined random projective plane drawing of Kn is
(n4/8π2)+O(n3). In analogy with the relation of Moon’s result to Hill’s
conjecture, Elkies asked if limn→∞crP(Kn)/n4=1/8π2.
We construct drawings of Kn in the projective plane that disprove this.