Thomas O'Brien's new algorithm helps coach Illini Hockey

3/3/2015 Ashish Valentine, ECE ILLINOIS

The newest addition to Illini Hockey's coaching staff is an algorithm developed by PhD student and assistant coach Thomas O'Brien to analyze goalies' weaknesses and build practice plans.

Written by Ashish Valentine, ECE ILLINOIS

Thomas O'Brien, middle left, supervising practice.
Thomas O'Brien, middle left, supervising practice.

An arena echoes with the sounds of razor-sharp metal scraping against ice, and the collisions of vulcanized rubber on wood. A goalie watches team members lining up pucks behind the fog of his breath as he prepares to block another shot. What’s interesting here is that it wasn’t his coach that set him on this drill: it was an algorithm. 

When its designer, ECE PhD student Thomas O'Brien Jr., isn’t fabricating lasers or teaching classes, he’s one of the assistant coaches of the Illinois men's hockey team. Ever since he took a series of data analysis classes two years ago, O’Brien has been keeping track of his performance using statistics, and then eventually collected those statistics to create an algorithm that makes judgments about his goalies’ strengths and weaknesses and proposes drills for daily practice. 

Thomas R O'Brien, Jr
Thomas R O'Brien, Jr

“At first, it was just something I kept in mind while I played,” O’Brien said. “When I started coaching, I was interested in keeping data on the team in a notebook while I watched, and I slowly started using the information to direct practice. If a goalie missed a lot of rebound shots from a certain angle, for instance, I drilled him accordingly.” 

The more O’Brien learned about manipulating and analyzing data, the more he built into a robust algorithm to analyze goalies on his team and make judgments about practice regimens. 

To use his algorithm, O’Brien first meticulously collects a number of statistics by watching film of the team’s past games. He collects data like the frequency of shots that goalies miss from particular directions, the location on the net that shots target, and where other players were on the ice.

When measuring a goalie’s performance, it’s important to look at the defending players assisting the goalie: for instance, an odd man rush shot where the offenders outnumber the defensemen puts a goalie in a very different situation from one in which he has much more help from the defense. 

Using this information, O’Brien calculates statistics on his goalies like save percentages, or the percentage of shots they’ve blocked. To get perspective on the goalie’s performance, he compares these percentages with averages for the team and the division. 

He then enters the statistics for each goalie into his algorithm, and the program compares each goalie’s performance with his peers to find out which areas the goalie needs to work on the most. Apart from comparing them with the average, it matches them up with a drillbook O’Brien has programmed in to develop a practice plan for each goalie, while taking into account their time constraints and other filters. 

The Illini hockey team lines up for a drill.
The Illini hockey team lines up for a drill.

For example, if the algorithm finds out a goaltender misses more odd man rush shots than the average, it could tell O’Brien to drill him more on blocking shots from several offending players. O’Brien noted that it was a mechanical representation of what he’d be doing as coach, anyway, but much more efficient. 

According to O’Brien, the hardest part is still manually collecting the data, as right now he has to sit through hours of footage to record statistics he wants to feed into his algorithm. 

“There isn’t yet a way to automatically collect these statistics, so often I’ll just analyze hours of film while we’re on the bus to an away game,” O’Brien said.  

Apart from coaching hockey, O’Brien studies under Associate Professor John Dallesasse. His research has him designing and fabricating vertical cavity lasers: beams of light that can be used to deliver data. His work involves scaling these lasers down to a size that could let them be embedded on silicon semiconductors, which would allow for integrated circuits based on light instead of electrical currents. 

John Dallesasse
John Dallesasse

After O’Brien graduates, he hopes to find a way to fuse his analytical aptitude with his love for hockey in the way that he has by working on this practice algorithm. 

“I know that I love hockey and applying the problem-solving skills I’ve developed as an engineer to coaching," he said. "If I could find a way to use all the tools I’ve learned while staying involved with hockey, that’d be perfect.” 


Share this story

This story was published March 3, 2015.