Disney Research System Predicts Soccer Goals

An orange-clad goalkeeper dives towards a soccer ball, while another player in a purple uniform watches
Photo by Alex Menendez/Getty Images
Goalkeeper Donovan Ricketts of Orlando City SC can't stop a shot by Mix Diskerud of New York City FC

Last weeked, Brazilian superstar, Kaká, scored in the final minute of his Major League Soccer debut for Orlando City Soccer Club. It was his fifth shot at goal, the second on target. Carlos Rivas, Orlando’s Colombian striker, also had five shots, but didn’t score. The match, against New York City FC, ended 1-1.

Soccer coaches often complain that their teams don’t take all the chances to score that they get. But, according to a group of Disney Research scientists, more chances don’t necessarily mean more goals. Germany beat Brazil 7-1 in the World Cup semi-final, last July, but Brazil had more shots on goal.

The Disney Research team released their paper, “Quality vs Quantity: Improved Shot Prediction in Soccer using Strategic Features from Spatiotemporal Data” at the 9th Annual MIT Sloan Sports Analytics Conference in Boston, on February 27th

They used player tracking data from sports analytics company Prozone looking at a season’s worth of shots on goal from an anonymous professional league. The data ccovered a ten-second window of play before each of the 9732 shots were taken.

During a soccer match, players are moving all the time. The role they play at any given moment will depend on the context of the match as it unfolds as much as their preassigned duties.  “We needed to align the tracking data so it told us which role a player is taking up, in each frame, rather than just their starting position,” says Patrick Lucey, who led the research team.

To work out the probability of different types of chances turning into goals, Lucey’s team used role representation software to analyse what players were doing, where and with who. “This enabled us to capture the nuances in general play, counter-attack, work out what role a player is performing during the 10 second segment being analyzed,” Lucey adds.

They also clustered each play into a specific match-contexts. This revealed that not all chances are created equal. The highest percentage of goals, they discovered, resulted from counter attacks—14.87 percent. Next came set pieces from a cross from a free-kick (10.05 percent), corners (8.97 percent) open play (8.26 percent) and finally free-kicks themselves (4.82 percent)

Nonetheless, two of Germany’s goals against Brazil in the World Cup came on the counter attack, one from a corner, and four from open play. Kaka’s debut goal was a deflected free kick. Analytics still can’t account for everything in the beautiful game.

Advertisement

Tech Talk

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.

Newsletter Sign Up

Sign up for the Tech Alert newsletter and receive ground-breaking technology and science news from IEEE Spectrum every Thursday.

Advertisement