Chapter 1 Introduction

Pace-of-play is an important characteristic in possession-based sports that can heavily influence the style and outcome of a match. In basketball, pace is defined as the number of possessions per 48 minutes (Ferrero, 2013). However, there is no standardized or generally accepted definition of pace in soccer. Pace has been defined as the number of shots taken (Knutson, 2013; Knutson, 2015) or the number of completed passes per game (Minkus, 2017). Both of these metrics can provide an idea of how fast the ball is moving. However, the main limitation of event-based pace metrics is their failure to appropriately account for the outcome or the circumstances under which they are performed. For example, evaluating pace as the number of shots taken does not account for the percentage of shots on target, while the number of completed passes does not differentiate between a pass made between two defenders in their own half of the pitch and a pass from a winger trying to create a goal-scoring opportunity. Pace-of-play has also been measured as the distance covered over time within a team’s possessions (Lawrence, 2015). However, short possession sequences do not provide an accurate measurement of pace. For example, a possession consisting of a goal kick and a pass may travel at a fast speed, but is not necessarily representative of a team’s overall pace.

The main goal of this work is to explore possessions via pass velocities. This new perspective of pace-of-play analyzes possessions that consist of three or more pass or free kick events, as these types of events are more definitive of a team’s pace. The use of spatio-temporal event data allows for more granular measurements of pace-of-play, such as measures of speed between consecutive events and between different regions on the pitch.

In addition, we aim to determine whether the pace metrics are useful in predicting the outcome of a match and whether those variables are significant.

Our research goals are three-fold:

1. Examine how pace-of-play varies across the pitch, between different leagues, and between different teams.

2. Quantify variations in pace at the league and team level and provide metrics to assess how well teams attack and defend pace.

3. Evaluate the effectiveness of the pace metrics by incorporating them into models that predict the outcome of a match.

The remainder of this paper is organized as follows. Section 2 describes the data and data pre-processing steps. Section 3 provides the framework and evaluation of our pace-of-play metrics. Section 4 discusses the modeling methodology and results. Section 5 includes a discussion of our findings and future work.