Three weapons, three regimes of order: per-touch skill transmission and the structural predictability of fencing tournaments
Year: 2026
Type: article
Abstract: Competitive predictability depends both on how rules resolve a contest and on how tournament structure converts results into rankings. Olympic fencing permits an observational comparison of three weapons that share a broadly common format but differ in scoring. We examined whether their ordering is consistent across pools, individual touches and elimination outcomes. We assembled 158,652 pool bouts from 474 International Fencing Federation Junior World Cup competitions, 2015–2026. The primary ability measure was pre-event FIE ranking points. Three analyses measured cyclic triads within pools, fitted independent-touch and co-scoring race-to-five likelihoods, and compared ability with the elimination round reached. Competition bootstraps quantified uncertainty, and an adaptive-quadrature hierarchical model described between-competition variation. Direct cross-round concordance and a signed-pair decomposition separated results from within-round seeding order. Pool-fitted slopes were also propagated into complete-data elimination tableaux. An equal-strength bracket simulation provided a structural reference for the recoded outcome. Cyclicity ranked épée, sabre and foil (14.5%, 12.4% and 10.3%), whereas per-touch transmission ran in reverse (β = 0.403, 0.527 and 0.645); all pairwise differences excluded zero. The co-scoring parameter was estimated at 9.34% in épée (95% CI 8.72–9.98) and zero in foil and sabre; under the fitted model it accounted for 36% and 71% of épée’s slope deficits, without eliminating them. Between-competition dispersion was largest in foil (σ = 0.205; 95% credible interval 0.177–0.235), though not on a scale-free measure. The pool ordering did not recur at tournament level, where a different ordering appeared: direct cross-round concordance was 0.694 in épée, 0.674 in foil and 0.645 in sabre, with the foil–épée contrast unresolved. Post-pool seeding partly determines within-round published classification, making their apparent agreement self-referential. Earned-tier recoding removes that ordering, and direct pair counting avoids an invalid equality between normalized Kendall correlation and cross-round concordance; the equal-strength bracket reference is nonetheless about a fifth of the observed recoded concordance. Under the fitted model, co-scoring accounts for part of épée’s lower per-touch transmission. All three weapons depart from their fitted two-route references, positively only in épée; final scores cannot identify temporal dependence. These observational findings support careful outcome definition rather than causal attribution.
Institutions: Manisa Celal Bayar University, Istanbul Topkapi University
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $2,890
OpenAlex ID: W7204473227