PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
June 14, 20264 citationsOpen Access

A Spectral Barrier for Paley Tournament Discrepancy, with an Application to a Maximum-Principle Refutation in Bradley–Terry Sigmoid-Tilt Models

View Full Paper
VRVladimir Riabov

Key Points

  • This research aims to explore the maximum principle for Bradley–Terry sigmoid-tilt models and demonstrate a counterexample in Paley tournaments.
  • Analyzed the maximum acyclic-subgraph advantage with respect to tournament direction.
  • Constructed an explicit, deterministic counterexample at the Paley tournament with 811 vertices.
  • Developed a verification script to reproduce numerical claims and certify bounds.
  • Found r(τ) < 2γ, directly refuting the maximum principle.
  • Demonstrated that every upper bound on g(Paley(p)) is at least (1/π)·p^(3/2)·ln p, proving log-free bounds are impossible.
  • Identified non-constructive refutation of the principle for n ≥ 787.

Abstract

We study the maximum principle for Bradley–Terry sigmoid-tilt models: the assertion that the base-2 log-moment-generating functional supᵤ Φₛ (u) of an independent Bradley–Terry product never exceeds the support function hLO (s) of the linear-ordering polytope on the tilt box ‖s‖_∞ ≤ 1. A single-point reduction shows that for a ±1 tournament direction τ the base point u = 0 gives an exact violation test: it violates the principle precisely when r (τ): = g (τ) / (n (n−1) /2) 550. 45 in g-units), the near miss at p = 787, the cotangent singular-value formula, the class-number cross-check Στ = p·h (−p), exact Held–Karp values for small Paley tournaments, and the first-moment crossover n = 787. It runs in seconds and requires only numpy and mpmath.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Vladimir Riabov (2026) studied this question.

synapsesocial.com/papers/6a2e4855b1cc60ccdea8ca12https://doi.org/10.5281/zenodo.20670429
Ask AI
Helpful
Bookmark
Share
View Full Paper