‘Moneyball’ twist: What’s a hockey player worth?

U. TORONTO (CAN) — What Billy Beane and Paul DePodesta did for baseball, an engineering duo hopes to do for hockey.

Portrayed in the movie Moneyball, Beane and DePodesta found success by bringing careful statistical analysis to bear on assembling the Oakland A’s roster.

At the center of their strategy was identifying the value of a player using nuanced metrics such as on-base percentage and slugging percentage. Timothy Chan, an industrial engineering professor at the University of Toronto, and undergraduate engineering student David Novati are bringing the same evidence-based sensibility to hockey.

In a paper soon to be published in the journal Interfaces—along with a follow-up paper presented at the 2012 MIT Sloan Sports Analytics Conference—Chan and Novati propose a new methodology for quantifying the value of a hockey player. A player’s value can determine not only his or her salary, but also influence trades and playing time.

The development of the model began in 2009 when the engineers started to contemplate how they could apply their knowledge to the Vancouver Winter Olympic Games.

“We started thinking about how we could develop a model that would help predict what the ideal Team Canada hockey team would look like,” explains Chan.

In the process, they realized a need for a better method to value different player types, which is the focus of their current research.

Chan explains: “Take Sidney Crosby as an example. Our model considers him a top-tier player because he gets a lot of goals and assists, and generally has a good plus-minus. He adds a lot of value to his team. Now you have other players who are more physical, deliver hits, and who get into a lot of fights. Our model says they don’t provide a lot of value to the team, contrary to what some people think.”

To arrive at that determination, Chan and Novati monitor a variety of player performance stats, including goals, assists, hits, blocks, time in the penalty box as well as time on ice. Their model considers a large number of statistics in concert to get an overall picture of a player.

In all, the model established four types of forwards (Top Line, Second Line, Defensive, Physical), four types of defensemen (Offensive, Defensive, Average, Physical), and three types of goalies (Elite, Average, Bottom).

The researchers continue to test and refine their model. Statistics for hockey, unlike baseball, are less plentiful. Eventually they hope to make their statistical model available to the NHL and to create an online tool for hockey fans.

More news from the University of Toronto: www.news.utoronto.ca/