Algebraic Degree of Network Games: Balanced Factors and Strategy Scaling

Abstract

The algebraic degree of a network game is the generic number of isolated complex-torus solutions of its polynomial indifference system. We recast the classical semi-mixed coefficient formula as an edge-marked player-level polynomial whose monomials are balanced directed multigraphs with prescribed in- and out-degrees. This representation separates the intrinsic counting problem from strategy-label refinements: exact evaluation remains $\#\mathrm{P}$-complete, but for a fixed number of players it is polynomial in the numerical strategy dimensions, while nonvanishing is decided by a capacitated flow test. It also characterizes inclusion-minimal positive-degree supports as integral transportation forests and yields a sharp $2N-1$-arc positive core. Exact-support coefficients form connected transportation fibers and give a nonnegative support calculus. Under proportional strategy growth $m\mathbf{k}$, a lattice local limit theorem expresses the first-order degree asymptotic through the capacity and maximum-entropy flow of the essential support. All counts are generic and complex; reality and simplex feasibility remain payoff-dependent.

28 pages, 1 figure, 1 table. Substantially revised and reorganized version; statements, proofs, exposition, and references have been extensively updated