Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(6.5), proof of Proposition 6.2, p. 26 — in a zero-sum game with an interior equilibrium x*, F_h(x*, y(t)) is constant along (RL)

Proved
ReinfRegGames.TimeAvg.display_6_5

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

conservation-lawfenchel-couplinglearning-in-gamesp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1zero-sum-games

Let G\mathcal GG be a 2-player game with finite strategy sets A1,A2\mathcal A_1, \mathcal A_2A1​,A2​ that is zero-sum, u1=−u2u_1 = -u_2u1​=−u2​, and let x∗=(x1∗,x2∗)x^* = (x_1^*, x_2^*)x∗=(x1∗​,x2∗​) be an interior Nash equilibrium: every pure strategy of every player has positive probability under x∗x^*x∗. Let each player use a penalty function hkh_khk​ (Definition 2.1), and let (y(t),x(t))(y(t), x(t))(y(t),x(t)) be an orbit of (RL). Write

Fh(x∗,y)=∑k=1,2[hk(xk∗)+hk∗(yk)−⟨yk∣xk∗⟩]F_h(x^*, y) = \sum_{k=1,2} \big[h_k(x_k^*) + h_k^*(y_k) - \langle y_k | x_k^* \rangle\big]Fh​(x∗,y)=k=1,2∑​[hk​(xk∗​)+hk∗​(yk​)−⟨yk​∣xk∗​⟩]

for the total Fenchel coupling. Then for every t≥0t \ge 0t≥0,

ddtFh(x∗,y(t))=⟨v1(x)∣x1−x1∗⟩+⟨v2(x)∣x2−x2∗⟩=0,(6.5)\frac{d}{dt} F_h(x^*, y(t)) = \langle v_1(x)|x_1 - x_1^*\rangle + \langle v_2(x)|x_2 - x_2^*\rangle = 0, \tag{6.5}dtd​Fh​(x∗,y(t))=⟨v1​(x)∣x1​−x1∗​⟩+⟨v2​(x)∣x2​−x2∗​⟩=0,(6.5)

and consequently Fh(x∗,y(t))=Fh(x∗,y(0))F_h(x^*, y(t)) = F_h(x^*, y(0))Fh​(x∗,y(t))=Fh​(x∗,y(0)) for all t≥0t \ge 0t≥0.

This conservation law is the key step of Proposition 6.2: a constant Fenchel coupling to an interior point keeps the players' score differences bounded (Proposition C.5).

Formalization Note The derivative at t=0t = 0t=0 is the right derivative. Players are indexed by Fin 2 (player 1 is index 0, player 2 is index 1), and u1=−u2u_1 = -u_2u1​=−u2​ is u0(s)+u1(s)=0u_0(s) + u_1(s) = 0u0​(s)+u1​(s)=0 for every pure profile sss. Both hypotheses, zero-sum and interiority of x∗x^*x∗, are used: after u2=−u1u_2 = -u_1u2​=−u1​ the middle expression of (6.5) equals u1(x1,x2∗)−u1(x1∗,x2)u_1(x_1, x_2^*) - u_1(x_1^*, x_2)u1​(x1​,x2∗​)−u1​(x1∗​,x2​), which vanishes because at an interior equilibrium each player is indifferent among all of its strategies against the other's equilibrium strategy.

Preamble
import Mathlib
import Definitions.Def_agt_games
import Definitions.Def_ReinfRegGames_Extinction_Model
Formal statement
namespace ReinfRegGames.TimeAvg

theorem display_6_5 {A : Fin 2 → Type*} [∀ k, Fintype (A k)] [∀ k, DecidableEq (A k)]
    (u : Fin 2 → (∀ k, A k) → ℝ) (hzs : ∀ s, u 0 s + u 1 s = 0)
    (xstar : ∀ k, A k → ℝ) (hNash : AGT.IsMixedNash u xstar) (hint : ∀ k α, 0 < xstar k α)
    (h : ∀ k, (A k → ℝ) → ℝ) (K : Fin 2 → ℝ) (hpen : ∀ k, ReinfRegGames.Extinction.IsPenalty (h k) (K k))
    (y x : ℝ → ∀ k, A k → ℝ) (horb : ReinfRegGames.Extinction.IsRLOrbit u h y x) :
    (∀ t, 0 ≤ t → HasDerivWithinAt
      (fun s => ∑ k, ReinfRegGames.Extinction.fenchelCoupling (h k) (xstar k) (y s k)) 0 (Set.Ici 0) t) ∧
    ∀ t, 0 ≤ t → ∑ k, ReinfRegGames.Extinction.fenchelCoupling (h k) (xstar k) (y t k) =
      ∑ k, ReinfRegGames.Extinction.fenchelCoupling (h k) (xstar k) (y 0 k) := by sorry

end ReinfRegGames.TimeAvg
Source
Mertikopoulos & Sandholm, Learning in games via reinforcement and regularization, arXiv:1407.6267v2, p. 26, proof of Proposition 6.2, (6.5)
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me