Mirzakhani's recursion formula computes volumes of moduli spaces of hyperbolic surfaces with geodesic boundaries — a breakthrough connecting symplectic geometry, intersection theory, and random surfaces. The volumes are polynomials in the boundary lengths, with coefficients encoding intersection numbers on moduli space. But hyperbolic surfaces can also have cone points — singularities where the total angle differs from 2π.
Note (2026-11-12; read against the paper's full text): the paper's theorem is for hyperbolic surfaces whose cone points all have angles in (0, pi]. It says that for cone angles between pi and 2 pi a pants decomposition may not exist (citing an appendix of Anagnostou and Norbury), and that Mirzakhani's method depends on a pants decomposition, so the result is restricted for that reason; the essay's statement that the pants decomposition does not care whether the special points are boundaries or singularities leaves out this restriction. The paper also cites earlier work: Do and Norbury obtained the recursion as the cone angle tends to 2 pi, and Anagnostou and Norbury extended the volume polynomials to surfaces with cone points and no geodesic boundary; this paper treats surfaces with both boundaries and cone points. The paper's own abstract states the polynomial in the boundary lengths and the imaginary cone angles.
The volume is still a polynomial, now in both boundary lengths and cone angles (arXiv:2603.11785). Generalized McShane identities — which decompose the surface into pairs of pants — yield a recursion formula extending Mirzakhani's to the cone-point case.
The structural observation: the polynomial nature of the volume survives the introduction of singularities. Cone points are not smooth — they break the local geometry — but the global volume of the moduli space remains polynomial in the new parameters. The variables change (imaginary cone angles join real boundary lengths), but the algebraic structure doesn't. This is a rigidity result disguised as a generalization: the polynomial form is not a consequence of smoothness but of the recursive structure of surface decomposition. The pants decomposition doesn't care whether the special points are boundaries or singularities.