LiU: helping LHC keep a clean sheet

A collaboration between two LiU researchers and the Linköping ice hockey club LHC will make it possible to discover the strategy of the opposing team quickly. The system is almost ready to be brought into use. Visualisation of sports data may have a profound influence on how ice hockey is played.

 Crowded in front of the LHC goal.  PETER HOLGERSSON

 

Patrick Lambrix and Niklas Carlsson, professor and associate professor, respectively, in the Department of Computer and Information Science, are both hockey fans. Patrick Lambrix has been a member of Linköping Hockey Club (LHC) for several years, while Niklas Carlsson plays in the club’s veteran team. The combination of experts in data analysis and hockey buffs led them to a collaboration with LHC.

“I often go to LHC’s matches and I started to wonder if we could help the team in some way”, says Patrick Lambrix.

They contacted the club at the end of 2017 and are now running a project on the visualisation of data. The data in this case are statistics of player actions during a match: shots on goal, passes, etc.

“The visualisation system that our group has developed allows Micke, who coaches the LHC goal tenders, to display an image of where all shots come from. This information can then be filtered in various ways. Maybe you only want to see shots taken when the attacker was close to the goal tender. Or to see which players were on the ice when the shot was taken. The coach can show the goal tender that these players often shoot from these positions”, Patrick Lambrix explains.

 

Revealing the opponent's strategy

The LHC goal tender coach Mikael Vernblom categorises everything that happens during the match. “The programme allows us to rapidly create images that reveal the match strategy of our opponents. I can show this to the goal tender in break periods during the match itself, and I can also describe before the match how the opponents play”, says Mikael Vernblom.

The system has not yet been taken into use, but soon will be.

 

Patrick Lambrix and Niklas Carlsson. Photo credit Magnus Johansson

 

Sports Analytics

The LHC project is an example of something known as “sports analytics”. The concept was introduced to the public through the Moneyball film in 2011. It shows how the Oakland Athletics baseball club used statistics to discover undervalued players. 
Patrick Lambrix and Niklas Carlsson are working on several similar projects.
“The visualisation with LHC is technically rather simple, but we are also using data to evaluate players and match strategy”, says Niklas Carlsson.
Traditionally, the ability of a player has been assessed by looking at such things as goals and assists. 

“One criticism of such a measure is that it doesn’t take the match situation into account. Scoring when your team is 5-0 up is not such a big deal, but if you’re 0-1 down, then scoring to equalise to 1-1 is extremely important. We have worked to include the match situation in the analysis”, says Patrick Lambrix.

The model developed by the LiU researchers, which is based on previous work by a Canadian group, looks at factors such as whether the team is ahead, and the period in which events occur. 
“We then use machine learning to analyse players and situations. The machine learning allows us to give a value to each action. These are then added together for any particular player and the higher value, the better the player”, Niklas Carlsson explains.

 

An image from the visualisation project with LHC. Certain information has been edited from the image.

 

Large amounts of data

Sports analytics are used in several sports, but work best in sports that produce large amounts of data. If goals or points are used as a measure, baseball and basketball are easier to work with than football and ice hockey.
“Another significant factor is whether the action is flowing or episodic. In football and hockey the game continues without interruption, which makes predictions difficult”, explains Patrick Lambrix. 
Predicting the motion of a football player in different situations is an example of such a difficult problem.
“We are working on this together with Signality, a company here in Linköping”, says Patrick Lambrix.

Even though this type of data analysis in ice hockey is common in North America, it is relatively new in Sweden. 
“We don’t know much about its use here, but I believe that other teams in the Swedish ice hockey league use it. But we don’t know what they do. And we are not aware of any other researchers who are working in a collaboration of this type”, says Patrick Lambrix.

He is teacher on a doctoral course in sports analytics, and hopes that it will be given as a single-subject course, starting in 2021.
Sports analytics is not always about improving player performance. Patrick Lambrix and Niklas Carlsson are discussing a project with the LHC doctor to use machine learning to learn how to avoid sports injuries. 

 

Where do you believe the project to evaluate players and match situations can lead? 

“If we are able to predict matches and correctly assess players, I believe that it will have a huge effect on how ice hockey is played”, says Patrick Lambrix.

Translated by George Farrants

 

Contact
Show/Hide content

Latest news from LiU
Show/Hide content