Ozan Candogan, Ishai Menache, Asuman Ozdaglar and Pablo A. Parrilo (2011), Mathematics of Operations Research 36(3):474–503 — DOI-10.1287/moor.1110.0500, ARXIV-1005.2405

Key takeaways

  • A game is a flow. The game graph has a node for each strategy profile and an edge between two profiles that differ in one player’s strategy. The deviating player’s payoff difference is a flow on it. The flow discards nonstrategic information but keeps “all relevant strategic aspects (e.g., equilibria)” (p. 1). Games with the same flow are “strategically equivalent” (p. 7).
  • Helmholtz decomposition (Theorem 3.1, p. 11). Any edge flow splits orthogonally into a gradient part (globally consistent), a harmonic part (locally consistent, globally inconsistent) and a curl part (locally inconsistent). A game’s flow has no curl part, because “The flow representation of a game defined in terms of utility functions (as opposed to preferences) does not exhibit local cycles” (p. 1), so its inconsistency is all harmonic.
  • Decomposition of games (Theorem 4.1, pp. 17–18). The space of games “is a direct sum of the potential, harmonic and nonstrategic subspaces”, with closed-form components and potential φ = δ₀†Du. The potential is the function whose gradient “approximates the flow in the original game (Du), in the best possible way” (p. 19). The decomposition is canonical: “the definitions of the subspaces do not rely on the inner product” (p. 19).
  • Potential and harmonic games (§5). “The set of potential games is equal to the subspace P ⊕ N” (Theorem 5.1, p. 22). Harmonic games have no potential component. In them, at a pure equilibrium “all players are indifferent between all of their strategies” (Lemma 5.1, p. 23), and so “Harmonic games generically do not have pure Nash equilibria” (Proposition 5.1, p. 23). The uniformly mixed profile is always an equilibrium of a harmonic game (Theorem 5.4).
  • Projection (§6). Under the weighted inner product (55), P, H and N are orthogonal (Theorem 6.1). “The closest potential game to G has utilities Πm φ + (I − Πm )um” (Theorem 6.2, p. 36). Its ε-equilibria bound the original game’s (Theorem 6.3).
  • Interpretation. “the potential component of a game captures interactions that can equivalently be represented as a common interest game, while the harmonic part represents the conflicts between the interests of the players” (abstract).

Standing in the record

Filed on 2026-10-23 at the owner’s request, as the formal ground for renaming “ascendancy” to “potential” (TERM-076, CLAIM-440) and for measuring it (QUESTION-268). Read in NOTE-1116, whose bearing section says where the reading fits and where it does not. The decomposition theorem is THEORY-894, with Jiang et al. (LIT-1439).