An Investigation into the Construction and Implementation of an xT Model in Rugby Union

: What is the accuracy of a predictive model based on historical data in predicting Rugby Union team performance?

Abstract

This thesis examines the notion of predictive models within the sporting industry; acting to explore, construct and discuss predictive models within Rugby Union. This project aims to build a model that acts to produce an expected score based on the events that occur within a game, that can be used to accurately project future scores or be used as an analysis tool. The model constructed applied the ideas and methods used in the development of expected goals models within football; to produce an effective model that utilises quantitative variables such as number of passes in a phase, alongside contextual variables such as the starting and ending location of a phase. The model demonstrated a baseline accuracy of 69.7% (and 73.8% when excluding outliers) when its scores were compared against the actual results; therefore demonstrating functionality. Several Pearson Correlation Analysis tests were completed, using several different predicted data sets compared to their actual counterparts; finding one significant, and numerous positive correlations. A slight limitation of this study was the reduced sample size, therefore warranting the need for further and future research into this area of performance analysis. However, through the exploration of this topic and therefore expansion of the academic knowledge, real-world application may be made more achievable and possible; examples of this may include: usage within the coaching industry or
even within the betting industry.

Date of Award	2026
Original language	English
Awarding Institution	Canterbury Christ Church University