Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proof of Proposition 5, p. 26 β€” at the true mean R, the capital process 𝒦ᡒ(R) is a martingale for Οƒ(L₁, …, Lα΅’)

Proved
RiskControl.WSR.K_martingale

by mikedeng1 Β· Oct 8, 2026 Β· Mathlib 0df444a (Lean v4.33.1)

betting-confidence-boundmartingalep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1risk-controlupper-confidence-bound

Let QQQ be a probability measure on R\mathbb RR concentrated on [0,1][0,1][0,1], let R=∫x dQ(x)R = \int x\, dQ(x)R=∫xdQ(x) be its mean, let n∈Nn \in \mathbb Nn∈N and δ∈R\delta \in \mathbb Rδ∈R, and let the sample Ο‰=(L1,…,Ln)\omega = (L_1, \dots, L_n)Ο‰=(L1​,…,Ln​) have the product law QβŠ—nQ^{\otimes n}QβŠ—n, so that L1,…,LnL_1, \dots, L_nL1​,…,Ln​ are i.i.d. with law QQQ. Let Ki(R)\mathcal K_i(R)Ki​(R) be the capital process of Proposition 5 and Fi=Οƒ(L1,…,Li)\mathcal F_i = \sigma(L_1, \dots, L_i)Fi​=Οƒ(L1​,…,Li​) (see the definition RiskControl.WSR.Process). Then the process

i↦Kmin⁑(i,n)(R),i=0,1,2,…i \mapsto \mathcal K_{\min(i,n)}(R), \qquad i = 0, 1, 2, \dotsi↦Kmin(i,n)​(R),i=0,1,2,…

is a martingale with respect to (Fi)iβ‰₯0(\mathcal F_i)_{i \ge 0}(Fi​)iβ‰₯0​ under QβŠ—nQ^{\otimes n}QβŠ—n; in particular, for 1≀i≀n1 \le i \le n1≀i≀n,

E[Ki(R)∣Fiβˆ’1]=Kiβˆ’1(R).\mathbb E\bigl[\mathcal K_i(R) \mid \mathcal F_{i-1}\bigr] = \mathcal K_{i-1}(R).E[Ki​(R)∣Fiβˆ’1​]=Kiβˆ’1​(R).

This is the martingale property at the heart of the proof of Proposition 5: since Ξ½i\nu_iΞ½i​ is Fiβˆ’1\mathcal F_{i-1}Fiβˆ’1​-measurable and E[Li]=R\mathbb E[L_i] = RE[Li​]=R, betting against the true mean does not change the expected capital.

Formalization Note QQQ is the law of the loss L(Y,TΞ»(X))L(Y, \mathcal T_\lambda(X))L(Y,Tλ​(X)) for the fixed Ξ»\lambdaΞ», which lies in [0,1][0,1][0,1] by the standing assumption of Β§3.1; the losses of the nnn i.i.d. calibration points are then i.i.d. with law QQQ. The paper's martingale is indexed by i=0,…,ni = 0, \dots, ni=0,…,n with F0\mathcal F_0F0​ trivial and K0=1\mathcal K_0 = 1K0​=1; it is extended constantly past nnn (index min⁑(i,n)\min(i,n)min(i,n), with Fi=Fn\mathcal F_i = \mathcal F_nFi​=Fn​ for iβ‰₯ni \ge niβ‰₯n), so that it is a process indexed by N\mathbb NN, as Mathlib's Martingale requires. The nonnegativity half of the paper's sentence ("non-negative martingale") is the separate item RiskControl.WSR.factor_nonneg. No condition on Ξ΄\deltaΞ΄ is needed: Ξ½i∈[0,1]\nu_i \in [0,1]Ξ½iβ€‹βˆˆ[0,1] is predictable for every real Ξ΄\deltaΞ΄.

Preamble
import Mathlib
import Definitions.Def_RiskControl_WSR_Process

open MeasureTheory
Formal statement
namespace RiskControl.WSR

/-- Proof of Proposition 5, p. 26 (arXiv:2101.02703v3): at the true mean `R = ∫ x dQ` of the
`[0, 1]`-valued losses, the capital process `𝒦_i = 𝒦_i(R; Ξ»)`, `i = 0, …, n`, is a martingale
for the natural filtration `β„±_i = Οƒ(L_1, …, L_i)` under the i.i.d. law `Qⁿ`; it is extended
constantly past `n` (index `min i n`) to a process indexed by `β„•`. -/
theorem K_martingale (Q : Measure ℝ) [IsProbabilityMeasure Q]
    (hQ : βˆ€α΅ x βˆ‚Q, x ∈ Set.Icc (0 : ℝ) 1) (n : β„•) (Ξ΄ : ℝ) :
    Martingale (fun i (Ο‰ : Fin n β†’ ℝ) => K n Ξ΄ Ο‰ (min i n) (∫ x, x βˆ‚Q)) (filt n)
      (Measure.pi fun _ : Fin n => Q) := by sorry

end RiskControl.WSR
Source
Bates, Angelopoulos, Lei, Malik & Jordan, arXiv:2101.02703v3, Proof of Proposition 5, p. 26, first display and the sentences around it
Human review
  • Endorsed by Shuze Chen Β· Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 Β· Oct 8, 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
Β© 2026 Prove2Me