Joel Quenneville took his cue for defensive pairings from statistical analysts. If you are sentimental toward the NHL’s current state, commit these games to your digital archive. Five years from now, they will be unrecognizable.
In hockey, progress happens regularly. Players become bigger, faster, and stronger. Equipment gets lighter and more durable. Coaches devise better game plans.
But the NHL will hinge — and change — upon the explosion of hockey intelligence. We are experiencing the game’s IQ transformation. Teams like Los Angeles and Chicago are mining data diamonds and applying them to the ice.
They understand that four lines of skill, speed, and puck-possessing prowess overwhelm the traditional template of two skilled units, a checking three some, and an energy group. They acquire and play mobile, pace-pushing blue liners over hold-your-ground defensive defensemen. They don’t panic when they fall behind, 2-0, because statistics show that scoring next impacts the outcome more than allowing a third goal.
Duncan Keith is Chicago’s best defenseman. But Keith sits against first lines. Chicago deploys Niklas Hjalmarsson in a shutdown role. That allows Keith to flourish in situations that play to his strengths: quickness, speed with the puck, accurate passing, and offensive instinct. Joel Quenneville didn’t conclude this based on what he knew about the game. The Chicago coach listened to his stats guys.
Data is everywhere. Wall Street banks piles of cash because they study companies’ price-earnings ratios and historical rates of return. Presidential campaigns research block-by-block polling results and voting records to target areas of improvement. Smart people created these world-changing innovations. They’ve applied their brains to other sports. In baseball, teams don’t bunt as often. More NFL coaches are going for it on fourth down instead of punting. The corner 3-pointer is the trend in the NBA.
Now the smarties are flooding the zone in hockey.
Edmonton hired Tyler Dellow, a former lawyer and stats-centric blogger, as a statistical consultant. New Jersey landed Sunny Mehta, an ex-professional poker player, as its director of hockey analytics. Eric Tulsky majored in chemistry and physics at Harvard and has a PhD in chemistry from Berkeley. Brian Macdonald was an assistant professor in mathematical sciences at West Point. Tulsky and Macdonald are consulting for undisclosed clubs.
Their employment is inspiring others. On Wednesday, during a panel at the Joint Statistical Meetings at the Boston Convention and Exhibition Center, speaker Sam Ventura , a PhD candidate in statistics at Carnegie Mellon, noted with a smile that he will be available for hire next year.
Ventura and other brainiacs, such as fellow panelists Macdonald, Kevin Mongeon (Brock University), Michael Schuckers (St. Lawrence), Michael Lopez (Skidmore), and Andrew Thomas (Carnegie Mellon), are already conducting groundbreaking work. They’re studying hockey’s granular events — odd-man rushes, zone entries, shots taken early and late in a shift — and uncovering information that complements intelligence gathered by traditional scouting.
Cracks exist in the latter method. One scout, for example, might prefer to conduct viewings on an AHL player on a Sunday, after he’s played road games on Friday and Saturday. The scout is trying to determine how the player performs when he’s tired to gauge his competitiveness. Another scout might ignore the Sunday viewing as garbage because of fatigue. This centers on a scout’s preferences.
Analytics is about sealing every crack. It’s not accurate enough to look at a goalie’s save percentage. It has to be adjusted for variables such as shot location, quality of competition, and save frequency. This requires math. The adjusted save percentage gives teams a more accurate depiction of a goalie’s skills.
The point, after all, is to find value. Teams regularly err by giving fat contracts to players who don’t deserve such plumpness (think four years, $13.5 million to Rob Scuderi). The smart clubs, with analytics as one of their tools, target players whose warts make them unworthy elsewhere. When Ottawa dismissed Ales Hemsky, Dallas signed him to a three-year, $12 million deal. If he plays with Jamie Benn and Tyler Seguin, Hemsky’s numbers will look like those balls tumbling from lottery machines.
The point is also to adjust game strategy and roster composition to mesh with the data. If teams board the analytics train, coaches will pull their goalies earlier to erase deficits. Forwards will leave a shot alone instead of blocking it. GMs will consider a roster with weight diversity — some 220-pounders with some 175-pound water bugs — instead of body uniformity. Coaches will not dress an enforcer who plays four minutes and chases the puck. Organizations will invest in forwards and goalies more than defensemen.
Such tweaks will require courage. They run counter to generations of hockey tradition. But people with lots of letters after their names will tell you that data doesn’t lie. The figurative data asteroid is expected to strike the NHL in 2015-16. By then, the league could introduce player motion tracking via SportVU, the company that performs the same service in the NBA.
Currently, stat geeks are tracking shot attempts, among other events, as a possession facsimile. If one team takes more shots (on net, missed, and blocked) than the other, it’s probably controlling the puck and creating more scoring chances.
The resulting stat is Corsi. It’s gaining traction among casual hockey observers. The analytics community would consider Corsi cute. These are guys who use phrases such as Poisson-like, Gaussian regularization methods, and shrinkage behavior (the latter being unrelated to Brad Richards’s Stanley Cup Final play).
Precise Corsi valuation depends on assuming the NHL’s real-time scoring system, which measures statistics such as shot location, giveaways, and hits, is accurate. It’s not entirely reliable. None of it is automated. At each rink, a group of off-ice officials tracks and logs these events from the press box — which, at some facilities, requires oxygen masks for entry.
Assuming the NHL welcomes player tracking, everything will be automated. The result will be clean, dependable data: a statistician’s dream. Smart people currently on the sidelines will sprint into the market.
“We’ll know where every player is on the ice at every moment,” Ventura said. “It gives us location information for all things happening over the course of the game — shots, hits, passes, people carrying the puck, where defensemen are positioned. Are they in the lane, ready to block a shot? Or are they giving up an easy lane for a goal. We can look at things like positioning for all plays, not just on shots. It’s really going to lead to a much richer data set.”
Player tracking and the information it provides won’t be the magic bullet. Unlike baseball, hockey isn’t a neat chain of static events. Players play offense and defense simultaneously. Substitutions happen on the fly. Goals aren’t scored regularly. A winger can be just as critical to a faceoff win as a center. Teams will need smart hockey people to eliminate the statistical noise.
People in the food industry are familiar with the bliss point, the perfect combination of salt, sugar, and fat. There will be something similar in hockey that marries good players, good coaching, and good analysis. The dinosaurs that miss this intersection will become extinct. The organisms that evolve and connect will thrive.
Perhaps above all else, analytics gets people thinking. Good information validates some theories. It nixes others. But it prompts us to consider ideas we might otherwise classify as foolish. In hockey, that’s a neat and novel approach.
Penn State grad has lowdown on scoring Hockey goes through a regular cycle of prospects seeking professional opportunities. The analytics community is no different. Students, even undergraduates, are crunching numbers, exploring hypotheses, and arriving at data-driven conclusions. Samantha Key, a 2014 Penn State graduate, was a presenter at the Joint Statistical Meetings.
Key, a Penguins fan, studied the importance of scoring first in the NHL. Key tabulated results from over 12,000 games between 2002-12. Key learned that teams that score first in the opening period won 66.7 percent of their games. Teams making it 1-0 in the second period won 69.1 percent of their games. The results spiked in the third period. Teams scoring first in the final period won 79.4 percent of the time.
Some of the conclusions Key teased out from her results:
■ Fatigue, injuries, and time are factors in the third-period jump. It’s harder for a tired team, maybe one featuring several hobbled players, to rally with less than 20 minutes to play.
■ It’s OK to enter first or second intermission down by one goal. It’s important for coaches and players not to panic in these situations. Chances of a comeback after 40 minutes are still better than if you fall behind at some point in the third.
■ If you have a 1-0 lead sometime in the first two periods, it’s not safe. It’s too early to ease off the offense and play conservatively.
■ If the game is scoreless after 40 minutes, it’s critical for coaches to think offensively at the start of the third. Skilled players deserve more ice time.
“You still want to play hard, but it’s not a big deal if you lose the lead. There’s still plenty of time,” Key said of the first two periods. “In the third period, it matters. That’s when if you don’t score first, you’re most likely going to lose.”
According to Michael Schuckers, one of the hockey analytics speakers at the Joint Statistical Meetings, off-ice officials at TD Garden undercount blocks and missed shots compared with the rest of the league. The result is a rink effect that skews the real-time data. It’s why teams prefer to track their own statistics in search of more accurate data. But this is an area in which player tracking will help players, too. Agents use the NHL’s real-time data in contract negotiations. For example, if Dennis Seidenberg is blocking more shots than he’s credited for, that could factor in his next deal."
Just a fad, it will never be adopted.