Visual Analysis of Soccer Match Using Player Motion Data

Since soccer data is based on the movement of twenty two players, soccer match data contains massive movement information of players and consists of complex properties and structures. Especially, regarding the fact that players move in the field constantly during the game, both spatial complexity and time complexity are high. Meaningful analysis of such complex data is able to help in figuring out characteristics of the team and establishing appropriate strategies for the team [1]. Also, patterns of attacks and defenses can be seen through players’ movement data as well, which serve as the key in understanding semantics of the game. Therefore, it is important to understand intuitively these data and analyze them contextually [2].

Hence, we propose a system that visually represent the movements of players during a match, by comparatively visualizing the different movements of players during attacks and defenses and the differences according to the formation or positions of the player such as 1 vs1 marks. Our system is as follows. First, we structuralize the player movement data and create metadata per position. Next, we visualize 2D movements, divide them into offensive and defensive patterns, and provide a user interface so that the user is able to select which information to visualize. Lastly, we determine the most frequently occurred information area by dispersion pattern of players based on the metadata, draw the determined area for each player, and assign distinct colors to these areas depending on the team of players. This enables comparative visualization.

Sample data are extracted from whoscored.com [3] which are recorded by very famous English professional soccer analyzers to design and develop a soccer game data visual analysis system. First, data scraping method and direct record methods are used to classify data as characteristics for design and removal of unnecessary data during data extraction process. Collected data are structured through data management system construction; then, the data are saved. By considering formation which consists team and position of players, player movement position in space of stadium is categorized and set as metadata. Then, data are selected for visualization and re-arranged due to that. Finally, it will be saved into a file as CVS format to read arranged data for development of visualization. These segmentation methods make users to recognized information from visualization more intuitive [4].

We suggest a visualization analysis system that analyzes game content based on movement data of players. Motion data will be the core of figuring out game content since it shows patterns of offense and defense. Each meta data is created by each position of players from structured data. Each data is expressed as distribution graph by categorizing the data based on the formation of a team. Offense direction of the team is set to increase understanding of game content analysis. Movement and location information functions from each position area are developed for comparable other players’ movement data. These can make to figure out player movement data easily even the formation is not matched and convenient to compare players’ movement when the strategy of each team is different. From Fig. 1, visualization due to team, player and positions are created by configuring control panel (step. 1). Visualization image data (step. 2) which corresponds to relative factors are constructed. Finally, required game content data are visualized from the request of users to be analyzed (step. 3).

Match content visualization analysis system procedures

Full size image

It is necessary to figure out the formation of players for each team to express the movement of players. 2D movement location data are figured out and saved into patterns of offense and defence by categorizing them. Users select optional data to visualize the movement of a player due to offense and defence of the player. Data forms are decided by player movement frequency location and location which are gathered as coordination of x, y in the 2D soccer field. By categorizing these forms of data, players and positions are constructed and confirm offense and defence patterns by each team. Offense and defence pattern are expressed by visualizing location value and movement for each position in each team’s base. The color of the team is expressed as #ffcc66 for A team and #ff9933 for B team. High-frequency location of players is colored into more primary color. If it is not, it has the color of 10 % saturation. As a result, the density of data, in other words, when the location of players’ movements is more duplicated into a certain location, the location will be closed to 100 % saturation of color. If not, it is expressed as minimum 10 % saturation. Visualization expressed into the corresponding color of team based on the most occurred information field by changing the movement of players into distribution form. We changed saturation value based on one color to compare more than two data property patterns of teams so that it is possible to compare with two patterns (Fig. 2).

Comparison methods for each player’s position

Full size image

Actual game data are used to deduce result from visualization analysis system and compare the movement of players and movement of defence due to the formation of A team and B team were compared. A team had 4-4-2 formation, and B team had 2-3-1 formation. The most important role of defence from soccer game is lowering instability to defence core regions [5]. The result of the defense line visualization is shown in Fig. 1(b). At the moment when the user selects a player or position, movement data is formed with use of the meta-data as shown in the Fig. 1(a) and the result is visualized Fig. 1(b). In our result the both teams use different formations, A teams plays in a 4-4-1 formation and B team plays 4-3-2-1 formation. Although both teams use different formations the use a back-four based defense line and that is why it is easy to compare and analyze our visualization result. In the enlarged image (b) different states of player distribution for both teams. Players in team A are more widely distributed and team B players are concentrated in the center. By analyzing player movement we can say that team A uses a more man-to-man defense tactic since the movement is widely distributed, while team B uses a defense tactic for a particular player. Our system can analyze offense as well. From the result of analyzing FIFA world cup in 2010, successful teams often conducted more side attack rather than center attack [2]. Our visualization analysis system can be used as a tool to confirm the correlation between a team winning and offense from a certain location. Of course, it can’t decide winning team from these analysis result. However, it is possible to figure out characteristics such as offense tendency and defense tendency of the team through this data analysis. Therefore, it is possible to bring good result by increasing the efficiency of a team from training players based on setting strategy which can increase the efficiency of team formation and encourages the potential of players into maximum [6] (Fig. 3).

Comparative visualization analysis by formation of A and B team

Full size image

By visualizing variety and complex soccer match data, it is easy to access information. So we designed and built a system which is convenient to a user by providing visual analysis. Only necessary data are primarily extracted from a large amount of game data to structure and arrange data to be accessed easily through the interface. Based on converted data, movements of players during a game are compared through visual analysis interface to conduct an experiment. The suggested visualization analysis system is not enough to implement in real. So, it needs a supplementary process by applying various visualization methods and analysis methods, as well as evaluation by experts. If the system is continuously developed through visualization methods of other information which is related with winning of soccer including data about player movements through modification and supplement process, the system can be applied to other team sports including soccer. Furthermore, it is possible to expect that this visualization analysis system can be provided as content to figure out the flow of games for public, as well as professional game analyzers.