Data analysis of passing style in football

I am a big football fan. I am also a researcher, so when I see those two topics intersect it is always a great pleasure. In a game that is constantly changing with influx of money, playing styles and new technology and big data study for games, it is of no surprise that an analysis aimed at the possibility of recognizing unique passing style from networks of passes is of big interest.

Usually, we focus on the statistical side of passing- success rate, number of passes, percentage of ball possession, etc, or the connection between pairs of players.

The published research “Searching for a Unique Style in Soccer” concentrates instead on network motifs – a local property of networks that focuses on recurrent and statistically significant patterns, and more specifically on passing patterns, through which the scientists could discern specific passing styles from different teams. Helpful from the network motifs is that usually they are no more than 3-4 nodes, usually the amount of people involved in a passing sequence during a game.


From paper: Searching for a Unique Style in Soccer at

The network motif, while usually static, here is used dynamically. The methodology of the analysis starts by extracting the passing sequence (for 3 consecutive passes), followed by computing the k-pass-long motifs. Furthermore, the number of passes is calculated, as well as the specific players involved and how they organize the flow of passes. The research uses data from the 2012-2013 football season from the Spanish, Italian, English, French, and German first divisions.

What the analysis discovered was that with 3 pass long motifs and a maximum time between passes 5 sec, there are five distinct passing styles emering: ABAB, ABAC, ABCA, ABCB, and ABCD. Nevertheless to prove those styles are correct, the investigators used comparison with a random pass network (generated 1000 times for each original pass network) with identical properties in terms of number of vertices and degree distribution.

Additional analysis to establish the similarities and differences in the motif characteristics of team is done through cluster analysis with k-means and hierarchical clustering method.

During the research, it was determined that the constant, tiring and sometimes dull passing from the Barcelona players (a style known as tiki-taka and practices now by Bayern Munich) has a distinct structure underlying the style and proven by using k-means and hierarchical clustering and both types showed that Barcelona’s styles fits in a cluster of its own. (Hardly surprising anyone who watches football). Even when the data for Barcelona is compared to that of other European football teams, Barcelona passing style for 2012-2013 season remain highly distinct. Such an observation is surprisingly made for the Italian club Torino, who furthermore shares similar style with Lile, Milan and Juventus, even though Torino was close to relegation that season.

It should be mentioned that this is still a preliminary work and more research has to be done in areas like various motifs and styles in home and away games, different player involvement in different motifs, as well as analysis of pass motives based on the final outcome of the games.