Source
Omidshafiei et al., LIT-1516, read in NOTE-1156 from arXiv v4: Theorem 2.4.24 (p. 14, proof p. 44).
What is shown and what is not
The paper argues this. The interior replicator flow preserves volume after a change of coordinates, so an asymptotically stable set must meet the boundary. Inductively it contains a vertex, and then the set of profiles reachable from that vertex by weakly better replies, which is an MCC. The change of coordinates is easy. In log-ratio coordinates the interior field has divergence 0, since no population’s payoff differences depend on its own mix, and a script at the status review of 2026-10-30 confirmed it. What is not carried out is the induction: that the trace of an attractor on a face is again an attractor there, and that the vertex’s weakly-better closure is pulled in. THEORY-942 holds the other half, that each MCC lies inside one chain component, re-derived. The converse of this entry, that every MCC lies in an asymptotically stable component, is false: in matching pennies the whole square is one component (NOTE-1156).