Modeling soccer (and perhaps your work team) as a social network

Frequent readers will be aware that an occasional hobby of mine is to numerically model sports and other interesting activities yet I lack much of the time or tools to do so. I have taken on NFL playoff football, both the entire playoffs and detailed modeling of the Superbowl down to the player level. I have also attempted, unsuccessfully, to model March Madness, the NCAA basketball playoffs, performing mostly analysis as opposed building a model that helps me win a March madness pool. Thus, I love to read interesting modeling papers, especially those which model sports or games as models for other real world activities.

The contributions of individual players in sports like football, baseball and basketball are helped by the large amount to statistics collected and available for these sports. Thus the contribution for individual team members to the team success is easier to model. Science online has a report of some work done by Jordi Duch and other researchers, at Northwestern and in Spain, that attempts to model the contributions of soccer players to the success of their team.

They point out that soccer is a very fluid game compared to baseball, or football and that combined with the very low scores makes statistics like goals and assists insufficient to model the contribution of players to the performance of the team. They hypothesize that the passes and flow of the game leading up to the rare goals are important for determining the outcome of the game and they use networks to model this flow. Players are nodes in the network and the lines between the nodes, called arcs, represent passes. Much as a Facebook or Twitter can be modeled as a network with friendship and interactions or follower/following being the connections, soccer is a "social" sport.

They also include nodes for the goal and for shots wide of the goal. To each of these arcs the attach statistics and probabilities from the 2008 European football championship on play pass accuracy, and goal accuracy to the arcs. One could them follow the ball through this "ball flow" network to a goal, a miss or to the other team. Combined with more calculations the group attempts to predict the outcome of soccer games.

Even more interestingly, the authors apply this concept to a work team that is writing a paper with several co-authors. Instead of the nodes being soccer players in paper network, a node represents a co-author in the manuscript, and the lines between the nodes represent communications directed from one co-author to the others. The e-mails represent communications between coauthors and the effectiveness of the authors is measured by completion of tasks like performing a calculation or scheduling a meeting. In the diagram below, author A2 (I think A3 in the second chart is a typos) seems to be an important and strong contributor.

One of the authors, comments on how the scheme can be used to assess the contribution of individual team members.

"One of the issues with any kind of teamwork is assigning the right credit," says Amaral. "The wild, loud people get more credit, but with this analysis you can get a picture of how much an individual really contributes to an outcome."

As work continues to evolve to be more team driven and highly networked, perhaps a scheme like this can not only point out strong contributors to a team, but also help an entire team work at a higher level. Imagine it applied to the work of developing open source software or Wikipedia articles.

(via Science online, the paper at Public Library of Science, PLoS, figures above are from the paper can be found at this .pdf link)