Number-loving sports like baseball and basketball have plenty of individual statistics to measure a team and a player’s performance over time. Not so for the “beautiful game.” "In soccer there are relatively few big things that can be counted," said Luis Amaral, professor of chemical and biological engineering at Northwestern University. "You can count how many goals someone scores, but if a player scores two goals in a match, that's amazing. You can really only divide two or three goals or two or three assists among, potentially, eleven players. Most of the players will have nothing to quantify their performance at the end of the match."

Amaral and his colleagues at NU’s McCormick School of Engineering and Applied Science knew that there was plenty of information buried in the data about player movement, passing and timing, if only a logical model could be created to explain the chaos on the field. Just as we learned in Part 1 about network models in basketball, Amaral, a lifelong soccer fan from Portugal, used his knowledge of social network analysis in biological systems to create a model of soccer ball strategy.

"You can define a network in which the elements of the network are your players," Amaral said. "Then you have connections between the players if they make passes from one to another. Also, because their goal is to score, you can include another element in this network, which is the goal."

Diagram 1 (click to zoom)

Using detailed player event data from the Euro 2008 international tournament, the researchers mapped every pass for each of the 31 matches. By comparing the passing route that a team used to end in a shot on goal, certain patterns emerge that highlight the players that are most often involved in the build-up to a goal. Not only could Amaral predict success for the team but also assign value ratings to each player.

"We looked at the way in which the ball can travel and finish on a shot," said Amaral, who also is a member of the Northwestern Institute on Complex Systems (NICO). "The more ways a team has for a ball to travel and finish on a shot, the better that team is. And, the more times the ball goes through a given player to finish in a shot, the better that player performed."

In Diagram 1, take a look at the bold network connections representing the most used passing paths between Euro 2008 players (denoted by their uniform numbers). Just as in the basketball network paths, finding passing patterns that end with shots on goals can give clues to the most efficient ball movement.

Diagram 2 (click to zoom)

From their initial research paper that appeared in PLOS One, Amaral created a company, Chimu Solutions, dedicated to refining their network model by adding data from thousands of top flight games across multiple leagues. Their goal is to produce a single metric that measures a player’s value as defined by their contribution to shots on goal.

In fact, by the time Euro 2012 came around last year, the model was able to attach an average player rating for each team. Diagram 2 shows the table for Spain’s champion roster which most pundits would agree ranks the players in a logical order from top to bottom.

This focus on data analytics is the new standard for sports and those teams that accept the challenge to find athlete monitoring tools that reveal these trends and patterns will have the competitive advantage over those teams that do not. Indeed, with apologies to Sun, the network is now the sport.

Follow @metrifit

Back in the mid-90s, Sun Microsystems, the creator of the Java programming language, coined the marketing slogan “The Network Is The Computer.” They were describing the Internet of twenty years ago, which obviously has grown into every corner of our lives today, as being as important if not more so than individual computers. The idea that individual nodes of a network can’t succeed on their own but only through communications and coordination sounds a lot like a pre-game pep talk in the locker room about teamwork and passing.

For continuous play sports like basketball and soccer, the optimal flow of the ball across a connected network of players is critical to winning. It was only a matter of time before network scientists, who were also sports fans, offered their advice on how these in-game connections can be measured and optimized.

In this two-part series, we’ll first take a look at research done at Arizona State University (ASU) on basketball, then, in the next article, an analysis of soccer networking and player metrics created by an engineering professor at Northwestern University. In fact, we'll see that "the network is the sport".

In traditional basketball offenses, transitions up the court begin with an inbound or outlet pass to the point guard who then becomes the hub of ball movement until his team eventually attempts a shot. But is that the most ideal strategy from a network flow standpoint? Even though a team’s point guard may be very skilled, would a less predictable ball movement be harder to defend?

Jennifer Fewell, a professor in ASU’s School of Life Sciences in the College of Liberal Arts and Sciences and Dieter Armbruster, math professor at ASU, watched and diagrammed every offensive series from the first round of the 2010 NBA playoffs to build a network model for each team. Knowing the eventual outcome of the first round and the entire postseason, they were able to correlate network movement with wins and losses. ”We were able to come up with a hypothesis about strategy and then apply network analysis to that,” said Fewell.

Diagram 1 (click to zoom)

First, take a look at Diagram 1, which represents the cumulative network model for all 16 first round playoff teams. The width of the arrows indicates the number of times that this network path was taken across the the hundreds of plays analyzed by the researchers. The wider the arrow, the more often those two players connected with a pass. As you can see, starting an inbound play with the point guard (PG) is very common while rebounds typically start with the big men near the glass, centers (CN), power forwards (PF) and small forwards (SF).

Diagram 2 (click to zoom)

However, Diagram 2 shows the network patterns for four teams with increasing success in the playoffs, the Bulls who lost in the first round, the Cavaliers in the 2nd round, and the Celtics who lost to the Lakers in the NBA Finals. The Bulls relied heavily on their point guard, Derrick Rose, while the champion Lakers used a more distributed model spreading the ball around in their famous “Triangle Offense.”

In fact, this more unpredictable pattern of the Lakers and the Celtics, which Fewell labeled a team’s entropy, was directly related to higher winning percentages across the playoff teams.

“What that basically says is that the most successful teams are the ones that use a less predictable, more distributed offense and that connect their players more,” said Fewell. “Those were the teams that had actually hired more elite players and allowed them to work together.”

Their research was published in PLOS One.

Network models like these also help coaches evaluate players as part of a team in a way that pure stats such as points, assists and rebounds may not capture. This is especially true in soccer, where scoring is much more rare than basketball. In our next article, we’ll take a look at the work of Professor Luis Amaral of Northwestern University and a new soccer stat that he calls, “flow centrality.”

Follow @metrifit