Proof of Theorem 7 (Appendix M), p. 36 — Pr(z_j ≥ z_i) ≤ 2e^{−mΔ²/(2(n−1)²)}
ProvedForecastComp.IELF.prob_wins_ge_leconcentration-inequalityforecasting-competitionlimit-accuracyp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let , let take values in on , let the outcomes be independent with marginals , and let all reports lie in . Let be forecasters and with . Writing for the number of I-ELF event lotteries won by forecaster ,
where the probability is over the outcomes and the event lotteries.
Preamble
import Mathlib import Definitions.Def_ForecastComp_IELF_Model
Formal statement
namespace ForecastComp.IELF
/-- Proof of Theorem 7 (Appendix M), p. 36: if forecaster `i`'s expected score exceeds `j`'s by
at least `Δ > 0`, then under independent outcomes the probability that `j` wins at least as many
event lotteries as `i` is at most `2 e^{−mΔ²/(2(n−1)²)}`. -/
theorem prob_wins_ge_le {m n : ℕ} (hn : 2 ≤ n) (R : ℝ → Bool → ℝ) (hbdd : IsUnitBounded R)
(θ : Fin m → ℝ) (hθ : ∀ k, θ k ∈ Set.Icc (0:ℝ) 1)
(y : Fin n → Fin m → ℝ) (hy : ∀ j k, y j k ∈ Set.Icc (0:ℝ) 1)
(i j : Fin n) (hij : i ≠ j) (Δ : ℝ) (hΔ : 0 < Δ)
(hgap : Δ ≤ avgScore R θ (y i) - avgScore R θ (y j)) :
probWinsGe R θ y j i ≤
2 * Real.exp (-((m:ℝ) * Δ ^ 2) / (2 * ((n:ℝ) - 1) ^ 2)) := by sorry
end ForecastComp.IELF
Source
Witkowski et al., arXiv:2101.01816v2, p. 36, proof of Theorem 7 (Appendix M)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.