EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

3303. FIVB ranking: a misstep in the right direction

Invited abstract in session TB-16: Ranking in sports, stream OR in Sports.

Tuesday, 10:30-12:00
Room: 19 (building: 116)

Authors (first author is the speaker)

1. Leszek Szczecinski
INRS
2. Salma Tenni
McGill
3. Daniel Gomes de Pinho Zanco
INRS

Abstract

In 2019, Fédération Internationale de Volleyball (FIVB) introduced a new "power" ranking, which relies on the probabilistic model relating the teams' strengths to the obtained outcomes. While there are many rankings which already use such models, the FIVB ranking is the first official one which uses the explicit probabilistic model in multilevel games (that is, games with more than three results).

In this work, we explain/assess the model and the on-line ranking used by the FIVB, placing it in a general statistical framework. In particular, we describe the ordinal model used in the ranking, show why it cannot be used directly for on-line ranking, and explain the approximations adopted by the FIVB. We introduce analytical perspective which explain why the approximations are suboptimal and propose an alternative formulation which significantly simplifies the ranking. Our analysis is backed up by the numerical optimization, where we use the official FIVB games to fit and to optimize the models we consider.

We conclude that, while the use of the explicit probabilistic model for multilevel ordinal data is an interesting and refreshing change in the world of rankings, and should be a preferred ranking strategy, the resulting algorithm is too complex and suboptimal. We thus formulate a number of recommendations, which vary in the amount of required changes and the performance/complexity trade-offs of the resulting rankings.

Keywords

Status: accepted

Back to the list of papers