Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 11, p. 5 — RVU (α > 0, β > 0, γ ≥ 0) and stability κ give each player regret ≤ α + βκ²(n−1)²T

Proved
RegLearnGames.IndivRegret.theorem_11

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

no-regret-dynamicsnormal-form-gamesp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1regretrvu

Consider an nnn-player game in which every player has ddd strategies and utilities ui(s)∈[0,1]u_i(s)\in[0,1]ui​(s)∈[0,1], played repeatedly: w0,w1,w2,…w^0,w^1,w^2,\dotsw0,w1,w2,… are mixed profiles, and uitu_i^tuit​ is player iii's utility vector against wtw^twt. Let α>0\alpha>0α>0, β>0\beta>0β>0, γ≥0\gamma\ge0γ≥0 and fix a horizon TTT. Suppose that every player's play satisfies the RVU inequality: for every player iii and every v∈Δv\in\Deltav∈Δ,

∑t=1T⟨v−wit,uit⟩≤α+β∑t=1T∥uit−uit−1∥∞2−γ∑t=1T∥wit−wit−1∥12,\sum_{t=1}^T\langle v-w_i^t,u_i^t\rangle\le\alpha+\beta\sum_{t=1}^T\|u_i^t-u_i^{t-1}\|_\infty^2-\gamma\sum_{t=1}^T\|w_i^t-w_i^{t-1}\|_1^2 ,t=1∑T​⟨v−wit​,uit​⟩≤α+βt=1∑T​∥uit​−uit−1​∥∞2​−γt=1∑T​∥wit​−wit−1​∥12​,

and that the play is stable: ∥wit−wit+1∥1≤κ\|w_i^t-w_i^{t+1}\|_1\le\kappa∥wit​−wit+1​∥1​≤κ for every player iii and every t=0,…,T−1t=0,\dots,T-1t=0,…,T−1. Then for every player iii and every wi∗∈Δw_i^*\in\Deltawi∗​∈Δ,

∑t=1T⟨wi∗−wit,uit⟩≤α+βκ2(n−1)2T.\sum_{t=1}^T\langle w_i^*-w_i^t,u_i^t\rangle\le\alpha+\beta\kappa^2(n-1)^2T .t=1∑T​⟨wi∗​−wit​,uit​⟩≤α+βκ2(n−1)2T.

In words, if the players use algorithms with the RVU property that also move slowly, then each player's regret grows like βκ2(n−1)2T\beta\kappa^2(n-1)^2Tβκ2(n−1)2T; for optimistic FTRL, β=η\beta=\etaβ=η and κ=2η\kappa=2\etaκ=2η, so the growth is of order η3\eta^3η3, which is what the choice of step size in Corollary 12 exploits.

Formalization Note The paper assumes the RVU property on any utility sequence; here it is assumed only on the realized play, which is weaker as a hypothesis and makes the theorem stronger. The paper's parameters "α>0,β>0,γ≥0\alpha>0,\beta>0,\gamma\ge0α>0,β>0,γ≥0" are kept and Definition 3's β≤γ\beta\le\gammaβ≤γ is not required, since the proof does not use it. Stability is required for the movements t=0,…,T−1t=0,\dots,T-1t=0,…,T−1 that enter the sum. The vector ui0u_i^0ui0​ is the utility vector against w0w^0w0.

Preamble
import Mathlib
import Definitions.Def_RegLearnGames_TotalRegret_Setting
Formal statement
namespace RegLearnGames.IndivRegret

theorem theorem_11 {n d : ℕ} (u : Fin n → (Fin n → Fin d) → ℝ)
    (hu : ∀ i s, u i s ∈ Set.Icc (0 : ℝ) 1)
    (w : ℕ → Fin n → Fin d → ℝ) (hw : ∀ t i, w t i ∈ stdSimplex ℝ (Fin d))
    (α β γ : ℝ) (hα : 0 < α) (hβ : 0 < β) (hγ : 0 ≤ γ) (T : ℕ)
    (hRVU : ∀ i, ∀ v ∈ stdSimplex ℝ (Fin d),
      ∑ t ∈ Finset.Icc 1 T, (v - w t i) ⬝ᵥ RegLearnGames.TotalRegret.utilVec u (w t) i ≤
        α + β * ∑ t ∈ Finset.Icc 1 T, ‖RegLearnGames.TotalRegret.utilVec u (w t) i - RegLearnGames.TotalRegret.utilVec u (w (t - 1)) i‖ ^ 2
          - γ * ∑ t ∈ Finset.Icc 1 T, RegLearnGames.TotalRegret.l1 (w t i - w (t - 1) i) ^ 2)
    (κ : ℝ) (hκ : ∀ i, ∀ t < T, RegLearnGames.TotalRegret.l1 (w t i - w (t + 1) i) ≤ κ) :
    ∀ i, ∀ wstar ∈ stdSimplex ℝ (Fin d),
      ∑ t ∈ Finset.Icc 1 T, (wstar - w t i) ⬝ᵥ RegLearnGames.TotalRegret.utilVec u (w t) i ≤
        α + β * κ ^ 2 * ((n : ℝ) - 1) ^ 2 * T := by sorry

end RegLearnGames.IndivRegret
Source
Syrgkanis, Agarwal, Luo, Schapire, Fast Convergence of Regularized Learning in Games, arXiv:1507.00407v5, p. 5, §3.2, Theorem 11
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