Theorems 1–2 — the pure price of anarchy of the average social cost of linear congestion games is 5/2
ProvedCongestionPoA.AsymSum.pure_poa_sum_eq_five_halvesFor linear congestion games, the pure price of anarchy of the average social cost is exactly . Precisely, with latencies , :
- (Theorem 1) in every linear congestion game with finitely many players and facilities, every pure Nash equilibrium and every pure strategy profile satisfy
- (Theorem 2) for every there are a linear congestion game with players, a pure Nash equilibrium and an optimal pure strategy profile with and .
The first part bounds the price of anarchy by ; the second shows that the bound is attained for every number of players , so is the exact worst case. This is the "" entry for asymmetric games and the average social cost in the paper's table of results.
Formalization Note The bound is stated multiplicatively, for every pure strategy profile , which is equivalent to the ratio bound and avoids dividing by . In the first part, players and facilities range over finite types in the lowest universe. In the second part, excludes the all-zero latency game, where the equality is trivial.
import Mathlib import Definitions.Def_CongestionPoA_AsymSum_Model
namespace CongestionPoA.AsymSum
/-- Christodoulou and Koutsoupias, *The Price of Anarchy of Finite Congestion Games*, STOC 2005,
PDF p. 3, Theorem 1 and Theorem 2 together (the abstract's "the price of anarchy is 5/2" for pure
equilibria and the average social cost):
1. (Theorem 1) in every linear congestion game, every pure Nash equilibrium `A` and every pure
strategy profile `P` satisfy `SUM(A) ≤ (5/2)·SUM(P)`;
2. (Theorem 2) for every `N ≥ 3` there is a linear congestion game with `N` players, a pure Nash
equilibrium `A` and an optimal pure strategy profile `P` with `0 < SUM(P)` and
`SUM(A) = (5/2)·SUM(P)`.
**Formalization Note.** Linear latencies are `f_e(k) = a_e k + b_e` with `a_e, b_e ≥ 0` (Sect. 2,
§1.1). The bound is stated multiplicatively, for every feasible `P`, which is equivalent to
`PA ≤ 5/2` and avoids dividing by `opt`. In the second part `0 < SUM(P)` excludes the all-zero
latency game, where `0 = (5/2)·0` holds trivially. Player and facility types in the first part range over
`Type` (universe 0), with finitely many players and facilities. -/
theorem pure_poa_sum_eq_five_halves :
(∀ {ι E : Type} [Fintype ι] [DecidableEq ι] [Fintype E] [DecidableEq E]
(G : CongestionGame ι E) (A P : ι → Finset E),
IsLinear G → IsPureNash G A → IsProfile G P → sumCost G A ≤ 5 / 2 * sumCost G P) ∧
(∀ N : ℕ, 3 ≤ N →
∃ (E : Type) (_ : Fintype E) (_ : DecidableEq E) (G : CongestionGame (Fin N) E)
(A P : Fin N → Finset E),
IsLinear G ∧ IsPureNash G A ∧ IsProfile G P ∧
(∀ Q : Fin N → Finset E, IsProfile G Q → sumCost G P ≤ sumCost G Q) ∧
0 < sumCost G P ∧ sumCost G A = 5 / 2 * sumCost G P) := by sorry
end CongestionPoA.AsymSum
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.