Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1 (State Space Collapse), pp. 1331–1332 (with a.e. rationality and W* > 0) — equilibria are exactly liftings of workload equilibria

Proved
FragLOB.Collapse.theorem_1

by mikedeng1 · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

limit-order-bookp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1queueingrouting-equilibriumstate-space-collapseworkload

Consider the fragmented limit order book market with NNN exchanges: dedicated limit-order rates λi≥0\lambda_i \ge 0λi​≥0, optimizing limit-order rate Λ>0\Lambda > 0Λ>0, market-order rate μ>0\mu > 0μ>0, attraction coefficients βi>0\beta_i > 0βi​>0, and arbitrary effective rebates r~0,…,r~N\tilde r_0,\dots,\tilde r_Nr~0​,…,r~N​. Investor types γ\gammaγ are distributed according to a probability measure FFF on [0,∞)[0,\infty)[0,∞) with finite mean.

  1. (Forward.) Let π∗∈P\pi^* \in \mathcal Pπ∗∈P be a routing decision and W∗>0W^* > 0W∗>0, and suppose that (i) π∗\pi^*π∗ is an optimal solution of the aggregated problem
max⁡π∈P ∫0∞{π0(γ)γr~0+∑i=1Nπi(γ)(γr~i−W∗μβi)} dF(γ)(13)\max_{\pi\in\mathcal P}\ \int_0^\infty \Big\{\pi_0(\gamma)\gamma\tilde r_0 + \sum_{i=1}^N \pi_i(\gamma)\Big(\gamma\tilde r_i - \frac{W^*}{\mu\beta_i}\Big)\Big\}\,dF(\gamma) \qquad (13)π∈Pmax​ ∫0∞​{π0​(γ)γr~0​+i=1∑N​πi​(γ)(γr~i​−μβi​W∗​)}dF(γ)(13)

and (ii) ∑i=1N(λi+Λ∫0∞πi∗(γ) dF(γ))=μ\sum_{i=1}^N\big(\lambda_i + \Lambda\int_0^\infty \pi^*_i(\gamma)\,dF(\gamma)\big) = \mu∑i=1N​(λi​+Λ∫0∞​πi∗​(γ)dF(γ))=μ (14). Then (π∗,Q∗)(\pi^*, Q^*)(π∗,Q∗) is an equilibrium, where

Qi∗=(λi+Λ∫0∞πi∗(γ) dF(γ))W∗μβi(15).Q^*_i = \Big(\lambda_i + \Lambda\int_0^\infty \pi^*_i(\gamma)\,dF(\gamma)\Big)\frac{W^*}{\mu\beta_i} \qquad (15).Qi∗​=(λi​+Λ∫0∞​πi∗​(γ)dF(γ))μβi​W∗​(15).
  1. (Converse.) If (π∗,Q∗)(\pi^*, Q^*)(π∗,Q∗) is an equilibrium and W∗=β⊤Q∗W^* = \beta^\top Q^*W∗=β⊤Q∗, then (π∗,W∗)(\pi^*, W^*)(π∗,W∗) satisfies (i) and (ii).

Here an equilibrium (Definition 1) is a pair (π∗,Q∗)(\pi^*, Q^*)(π∗,Q∗) with π∗∈P\pi^* \in \mathcal Pπ∗∈P, Q∗∈R+NQ^* \in \mathbb R^N_+Q∗∈R+N​, β⊤Q∗>0\beta^\top Q^* > 0β⊤Q∗>0, such that for FFF-almost every type γ\gammaγ the vector π∗(γ)\pi^*(\gamma)π∗(γ) maximizes π0γr~0+∑iπi(γr~i−EDi(Q∗))\pi_0\gamma\tilde r_0 + \sum_i \pi_i(\gamma\tilde r_i - \mathrm{ED}_i(Q^*))π0​γr~0​+∑i​πi​(γr~i​−EDi​(Q∗)) over the simplex, and λi+Λ∫0∞πi∗ dF=μi(Q∗)=μβiQi∗/β⊤Q∗\lambda_i + \Lambda\int_0^\infty\pi^*_i\,dF = \mu_i(Q^*) = \mu\beta_iQ^*_i/\beta^\top Q^*λi​+Λ∫0∞​πi∗​dF=μi​(Q∗)=μβi​Qi∗​/β⊤Q∗ for every exchange iii.

The theorem reduces the NNN-dimensional equilibrium problem to a one-dimensional one in the workload: equilibria are exactly the liftings, through (15), of solutions of the workload problem (13)–(14).

Formalization Note. Three changes to the printed statement, each needed for it to be true and meaningful. (a) Individual rationality in Definition 1 is required for FFF-almost every type instead of every γ≥0\gamma \ge 0γ≥0: as printed the forward direction fails (for N=1N = 1N=1, λ1=0\lambda_1 = 0λ1​=0, Λ=2\Lambda = 2Λ=2, μ=β1=1\mu = \beta_1 = 1μ=β1​=1, r~=(−1,1)\tilde r = (-1,1)r~=(−1,1), FFF exponential with mean 1, the policy "post iff γ>ln⁡2\gamma > \ln 2γ>ln2" with W∗=2ln⁡2W^* = 2\ln 2W∗=2ln2 satisfies (13)–(14), and redefining it to "market order" at the single type γ=5\gamma = 5γ=5 keeps (13)–(14) but violates (10) at γ=5\gamma = 5γ=5). (b) W∗>0W^* > 0W∗>0 in the forward direction, and β⊤Q∗>0\beta^\top Q^* > 0β⊤Q∗>0 in the equilibrium: with Λ=μ=1\Lambda = \mu = 1Λ=μ=1 in the same example, W∗=0W^* = 0W∗=0 satisfies (13)–(14) but (15) gives Q∗=0Q^* = 0Q∗=0, where (4) is 0/00/00/0. (c) FFF has finite mean, so that the objective of (13) is a genuine integral for every routing decision. The paper's assumption that FFF has a differentiable CDF with continuous density (p. 1329) is not used by Theorem 1 and is dropped, so the statement here is more general. The rebate assumptions (r~0<0\tilde r_0 < 0r~0​<0, Assumption 1) are not imposed. Integrals ∫0∞⋯dF\int_0^\infty\cdots dF∫0∞​⋯dF are taken over R\mathbb RR against FFF, which charges no negative type. At an empty exchange the delay is W/(μβi)W/(\mu\beta_i)W/(μβi​), the value (12) assigns.

Preamble
import Mathlib
import Definitions.Def_FragLOB_Collapse_Model

open MeasureTheory
Formal statement
namespace FragLOB.Collapse

theorem theorem_1 {N : ℕ} (M : Market N) (hM : M.Standing) (F : Measure ℝ)
    [IsProbabilityMeasure F] (hF : F (Set.Iio 0) = 0) (hmean : Integrable (fun γ => γ) F) :
    (∀ (pi : ℝ → Fin (N + 1) → ℝ) (W : ℝ), 0 < W → IsAggOptimal M F W pi →
        AggBalance M F pi → IsEquilibrium M F pi (lift M F pi W)) ∧
      (∀ (pi : ℝ → Fin (N + 1) → ℝ) (Q : Fin N → ℝ), IsEquilibrium M F pi Q →
        IsAggOptimal M F (workload M Q) pi ∧ AggBalance M F pi) := by sorry

end FragLOB.Collapse
Source
Maglaras, Moallemi & Zheng, Operations Research 69(4):1324–1348 (2021), pp. 1331–1332, Theorem 1 (State Space Collapse), with Definition 1 (p. 1331)
Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 22, 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