Player Vectors: Characterizing Soccer Players’ Playing Style from Match Event Streams

Abstract

Transfer fees for soccer players are at an all-time high. To make the most of their budget, soccer clubs need to understand the type of players they have and the type of players that are on the market. Current insights in the playing style of players are mostly based on the opinions of human soccer experts such as trainers and scouts. Unfortunately, their opinions are inherently subjective and thus prone to faults. In this paper, we characterize the playing style of a player in a more rigorous, objective and data-driven manner. We characterize the playing style of a player using a so-called 'player vector' that can be interpreted both by human experts and machine learning systems. We demonstrate the validity of our approach by retrieving player identities from anonymized event stream data and present a number of use cases related to scouting and monitoring player development in top European competitions.

SoccerMix: Representing Soccer Actions with Mixture Models

“…Our goal is to capture the playing style of either a player or a team. As in past papers [7,10,16,17], our intuition is that playing style is tied to where on the pitch a player (or team) tends to carry out certain types of actions. Most playing style analysis techniques follow the same two-step approach:…”

“…Traditionally, most research has focused on the second step [5,7,8,16]. However, picking sub-optimal groups in the first step can introduce significant problems such as sparsity down the line.…”

“…However, picking sub-optimal groups in the first step can introduce significant problems such as sparsity down the line. In fact, many sophisticated data aggregation methods such as pattern mining [8] and matrix factorizaton [7] are often only used in step two to combat the problems introduced by the sub-optimal groups established in the first step.…”

“…One approach is to lay a grid over the field and proclaim two actions to be similar when they are of the same type and fall in the same grid cell [5,7,16,17]. However, this approach has three downsides.…”

“…Third, ideally we would like to group actions on additional attributes such as ball direction, but considering more attributes makes each action more unique, which increases the sparsity in the data. Hence, most approaches only include one or two attributes in their analysis [2,7,14]. Rarely do approaches consider three or more attributes [8,17].…”