Notice
This record is in review state.
Closed access
Author
Date
2024-03-25Type
- Bachelor Thesis
ETH Bibliography
yes
Altmetrics
Abstract
In soccer analytics, the Poisson-Binomial distribution is used to obtain
the probabilities of the possible match scores of a team. This is an
important distribution for analysts, because it allows for construction of
new descriptive metrics and further meta-analysis.
This work outlines the theoretical background of the Poisson-Binomial
distribution, hoists it into the context of soccer analytics, while also
providing content on the background of the game of soccer. The contribution is analysis of real-world data and assessment of algorithms
which compute the Poisson-Binomial distribution. These algorithms
include both deterministic and probabilistic computation approaches.
The algorithms are introduced, and their properties discussed, then
assessed on execution time and numerical stability.
The goal of this work is to provide the reader with a recommendation on
what algorithm to use to calculate the Poisson-Binomial distribution in
the context of soccer analytics. This should give an anchor point for analysts, which are concerned with the algorithmic efficiency and precision
of their computation strategy for the Poisson-Binomial distribution. Show more
Subject
Soccer Analytics; Poisson-Binomial; Computation; Performance MetricsOrganisational unit
09610 - Brandes, Ulrik / Brandes, Ulrik
More
Show all metadata
ETH Bibliography
yes
Altmetrics