Reply To: MIT Sloan Sport Analytics Conference Video

[hockeygod]

For now, speed of game outpaces analytic angle

Fluto Shinzawa, Boston Globe, March 25 2012

On March 13, with his team down, 2-0, to Philadelphia with five minutes remaining in regulation, Devils coach Peter DeBoer did something that would have startled most of his counterparts. During four-on-four play, he yanked goaltender Martin Brodeur and replaced him with forward David Clarkson, leaving the net empty.

The brassy move caught the attention of Michael Schuckers, associate professor of mathematics at St. Lawrence University. Schuckers doubles as the cofounder of Statistical Sports Consulting. In that position, he consults with NHL teams, using data to promote less-considered approaches. For example, signing a mid-tier goalie and a forward instead of focusing all resources toward an elite puck-stopper. Or that winning faceoffs isn’t as important as coaches believe it to be. Or pulling the goalie earlier.

Schuckers bases the latter proposal on a paper by David Beaudoin and Tim Swartz, based on data from the 2007-08 NHL season. The researchers concluded that in such a scenario as New Jersey’s (trailing by two goals with six minutes left), the losing club should pull its goalie far sooner than with 1:30 remaining, the traditional mark when coaches consider the maneuver.

“If you’re in a situation where you’re pulling your goalie, you’re obviously already in a spot where you’re likely to lose,’’ Schuckers said. “You’re down a goal or two goals. You’re likely to lose and get no points.

“The strategy, admittedly, is a risky one. But it’s one which can pay off. If you pull your goalie with 4-5 minutes left, yes, you’re probably still going to lose. But the chance you get some points out of the game improves drastically.’’

With 4:49 left in regulation of that Flyers-Devils game, Danny Briere scored an empty-net goal. Philadelphia won, 3-0. But Schuckers still believes the Devils made the right move.

“By giving yourself an extra attacker, you’re increasing the probability you’re going to score a goal,’’ Schuckers said. “Over the course of repeating that strategy in an 82-game season, it eventually pays off.’’

That scenario and its outcome illustrate the challenge Schuckers and his colleagues face when pitching their theories to hockey’s gatekeepers. Earlier this month, Schuckers participated in a panel at the MIT Sloan Sports Analytics Conference. Also on the panel were Bruins general manager Peter Chiarelli, Toronto GM Brian Burke, NBC analyst Mike Milbury, and former NHLer Tony Amonte.

Burke dominated the discussion, downplaying the significance of the oft-cited but irrelevant plus/minus statistic, which largely reflects whether a player is on a good team or a bad team. (See Jeff Schultz’s plus-50 in 2009-10 or Tim Gleason’s minus-11 in 2010-11.)

But Burke also emphasized that he prefers the traditional methods of scouting. He noted that he watched just one of Ryan Kesler’s shifts at Ohio State to decide that he would draft the center in the first round of 2003. Burke also acknowledged scouting Wayne Simmonds as a junior multiple times and not seeing anything to command his attention.

Like most in the game, Burke endorses the use of experienced bird dogs in pro and amateur rinks to evaluate players and determine who stands above the rest.

As for the statistics gathered and applied by Schuckers and others in analytics, Burke used the metaphor of a lamppost for a drunk: useful for support, but not for illumination. You could say there is resistance.

“You still have a culture of hockey that is very old-school,’’ said Schuckers. “There’s plenty of skepticism in hockey for the role that hockey analytics and statistics can play.’’

Schuckers acknowledges the hurdles analytics must overcome in hockey. The practice has become standard in baseball, where the sport’s rhythm makes it perfect for statistical analysis.

Hockey, however, is a chain reaction of events that unfold rapidly. Take, for example, Chris Kelly’s goal in the Bruins’ 8-0 thrashing of the Maple Leafs last Monday.

That Kelly scored on Toronto goalie James Reimer is a static piece of information, of little use.

But by breaking down the series of events that led to the goal, the Bruins could accrue some valuable data. The goal took place in the first period. The scoring chance, which started off a Luke Schenn turnover, started against the left-side boards. It happened on a one-man forecheck. The Leafs had just played D-to-D, with Jake Gardiner passing to Schenn. Kelly scored on his backhand.

By examining each component, the Bruins could, for example, roll out a specific forecheck for a certain stage of the game against certain opponents.

However, to break down a 60-minute game into hundreds of events requires more video tracking and analysts to follow each sequence. Neither the NHL nor its teams currently use enough resources to gather the raw data.

“We can go through the play-by-play file and the files of NHL records of events in a game,’’ explained Schuckers, referring to the real-time data, supplied by off-ice officials, that the league currently tracks. “Take the beginning of a game. There’s a faceoff to start the game. About 40 seconds later, we have a shot, a shot by the team that won the faceoff.

“We know they won the faceoff. We know, 40 seconds later, they took a shot. What we don’t know is all those things that happened in between. Did they win the faceoff, go back into their zone, and organize the attack? Or did they immediately go into the offensive zone and pass it around for 30 seconds?

“Not knowing those pieces of information makes some of the in-game stuff hard to do.’’

It will take several years for analytics to gain relevancy in hockey. The NHL would have to install additional cameras in every rink to capture each event more thoroughly. Hockey-savvy statisticians would have to track every sequence.

Even if enough data is collected and analyzed to uncover trends, coaches must embrace the information, then convince their charges to carry out certain tasks on the ice. In its current state of infancy, analytics projects to be more helpful for management in player evaluation than for coaches, because of the sport’s speed.

“You might have six things that analytics tells you,’’ said Schuckers, repeating an anecdote he heard from Celtics assistant GM Mike Zarren. “Four of them, you’re confident enough to take to the coach. The coach is confident enough to tell two of them to the player. If you’re lucky, the player will execute one of them.’’

One hand or two?

Under former coach Ron Wilson, Toronto defensemen were encouraged to keep both hands on the stick. That way, they could be stronger on pucks and more forceful at swatting away opponents’ sticks. The Bruins coaches teach their defensemen to lead with their sticks with one hand and aim for the puck. With their other hand, they can lean on the puck carrier. For example, if a forward is trying to swing wide, the defensemen are taught to reach with their sticks with one hand, which gives them better reach. If the forward then tries to pass, the defensemen has a hand free to attempt to swat the puck away. In tight, the one-hand approach also works. “You use one hand with stick on puck, then push the puck,’’ Johnny Boychuk explained. “Then you can push him at the same time with your other hand.’’

Dean
M.Ed (Coaching)
Ch.P.C. (Chartered Professional Coach)
Game Intelligence Training
"Great education depends on great teaching."