Your cart

Phasing into Analytics: How Hockey Teams Can and Should Use Analytics

This is a continuation of a series by Shayna Goldman analyzing the integration of analytics into hockey analysis. 

. . . . .

As analytics and enhanced statistics are becoming more deeply integrated into hockey analysis, they are converging with the traditional eye-test. While some analysts have embraced the additional insights analytics can provide, NHL teams are still grappling with the proper role of analytics in hockey analysis and managing a club.

To explore how teams can and should use analytics, a number of experts in the field were consulted:  Garret Hohl, CTO of the data and analysis company HockeyData and founding manager of the Hockey-Graphs blog; Prashanth Iyer, contributor at Winging It In Motown and Hockey-Graphs; Ryan Stimson, a writer at Hockey-Graphs; Corey Sznajder, a hockey researcher and analyst; and Carolyn Wilke, editor and site manager for Today’s Slapshot.

. . . . .

While a number of teams have incorporated analytics into their front offices, many are still struggling to balance the differing approaches in their decision making. Hohl acknowledges that this is a part of the process for those teams expanding to include analytics: “Analytics is simply the study of statistics, in what they mean and what they predict. I would think of perfection as a journey of improvement, not a destination. The way that I look at it, we are only just beginning that journey.”

The Florida Panthers are a team that has embraced analytics, but they are not fully committed as Stimson noted. “I think you saw some good player acquisition this offseason for Florida, but they still dole out plenty of ice time to their fourth line in places they probably shouldn’t. Acquiring talent is only part of the equation.” Sznajder said that teams may not be looking at the correct statistics, which could explain why players are getting minutes at inopportune times. “Every team probably looks at stats and tries to break them down as much as possible, but it’s possible they’re looking at stuff that’s totally meaningless when it comes to predicting future performance. We know stats like hits are mostly random and meaningless but there’s probably teams that are looking at them.”

Finding the appropriate balance between the perspectives of the traditional approach and analytics may be a struggle for some teams. Iyer indicated that the issue may stem from a lack of commitment from the organization. “A lot of teams have made analytics hires because other teams are doing it, but the organizational buy-in is missing.” Some teams have made the front-office commitment, Iyer said, citing the Florida Panthers, Arizona Coyotes, and Carolina Hurricanes. Teams like the Toronto Maple Leafs and Pittsburgh Penguins have not fully transitioned, but have made a stronger commitment than most other organizations.

But simply having an analytics department may not be enough for it to have an influence. According to Stimson, “you need to have the buy-in throughout the organization.”

Many teams utilized analytics before they garnered such public attention, largely through the work of hockey bloggers and writers who aided in the process of “discovering which stats are important and which ones aren’t, which led to a few getting consulting jobs with teams” Sznajder explained. “Every team definitely uses statistics, in one capacity or another, with tracking things that they wish to track. Using analytics though, is actually testing these things, testing out what they are worth rather than taking worth relative to what you subjectively view them to be worth,” Hohl added.

According to Wilke, there is a disconnect between front-offices and analytics. “I think, overall, analytics­­––data driven insights––are viewed favorably by most general managers.” However, most do not let analytics “actually drive decision-making processes, as most of these guys have been working in hockey for many years and are likely to trust their own experience, whether those are aligned with the data are not” Wilke continued.

Get The Latest NHL Tech Insights In Your Inbox!

“General managers may view [analytics] as a tool to get a comprehensive look at a player or team, or they may use it to simply confirm what they want to know,” Stimson said. “I think the eye-test is part of the equation, but people that harp on intangibles and suggest what we can’t measure should matter more than what we can measure, that’s mostly about: self-preservation: ‘I’m a hockey guy and I know what matters,’” he added. Iyer furthered this idea of teams, even those with distinguished analytics departments, signing players that challenge analytical evaluations, pointing to Toronto signing Matt Martin or Carolina re-signing Cam Ward.

Some teams are broadening their hockey analysis by incorporating analytics. In addition to the Coyotes, Panthers, Penguins, and Hurricanes, Wilke cited the Avalanche, Lightning, and Wild as teams that have been investing in analytics.

Others, though, have visibly regressed. Iyer pointed to the Oilers and Canadiens. “We saw Edmonton and Montreal part ways with key members of their analytics departments this summer and, in turn, made a couple of head-scratching moves. However, I don’t know how much either of these organizations bought in in the first place, to truly call it a regression.” Stimson noted the Canadiens and Oilers as well. “What counts is the process, and I’m sure there will be countless people who say those teams were right all along if the return of Carey Price carries the Habs to the playoffs or Connor McDavid does the same for the Oilers. You can’t confuse results as meaning you have a sound process.”

Wilke expressed her confusion with the Maple Leafs this past offseason. “They’ve seemed to take a bit of a step back from being on the cutting edge of analytics, though they’re almost assuredly going to have much better results this year.” As for teams that need the most help to improve (and should incorporate analytics), Wilke notes the Canucks and Ducks.

. . . . .

Some enhanced statistics may provide greater value than others, in general or for certain teams. Hohl cites shot differentials, such as Corsi, as important statistics “due to the market undervaluing players’ contributions in these areas. As the market shifts, the next advantage will be in some other areas that are currently undervalued.” Iyer also pointed to Corsi, “specifically 5v5 score-adjusted Corsi for% … because it has been shown to be one of the best predictors of team performance.” Overall possession and Corsi for and against are statistics Wilke looks to understand who is helping and hurting the team.

It is important to note that one specific statistic cannot be narrowly focused on since, as Wilke describes, “hockey is too nuanced a game to be summed up in one number.”

Sznajder explains that teams should look to zone entries and neutral zone play, “because we know entries correlate with outshooting your opponent and are a repeatable stat, so they’re something every team should be looking at to evaluate their play.”

“A team like the Senators has had issues of giving up a lot of shots, and it could stem from their play in the neutral zone, whether it’s from their own team giving up the blue line when defending the neutral zone or not entering the opponent’s zone with control enough,” Sznajder continued. Similarly, the 2013-2014 Buffalo Sabers struggled in the neutral zone––failing to carry the puck into the offensive zone and generate shots. That same season, Sznajder noted, the Toronto Maple Leafs were “a disaster in all three zones, allowing their opponents to carry the puck in often while playing mostly dump-and-chase.” In contrast, teams like the Chicago Blackhawks received a bump in offensive production from their dominant neutral zone play.

Stimson finds that shot and expected goal differentials are an important metric for teams to study, helping “form the baseline of player and team analysis.” Passing data is another area for teams to look at for team and player evaluation; “it’s more important than individual shots for player evaluation, and it can dramatically improve how teams analyze their style of play.” In order to score more goals, which all teams look to do, “you need the data that measures certain tactics that you can analyze to see how well those tactics predict scoring.” Looking at the shot assists and success of passes from certain areas can help a team understand how their system can be tweaked to create more scoring chances, whether it is changing the forecheck, the structured plays in the offensive zone, transitional play, or defensive zone coverage.  

Analytics can also provide superior insights to standard metrics like +/-. Hohl feels teams should not be focusing on +/-, which includes “all even strength goals, all goals with either goaltender pulled, and short-handed goals.” Because the +/- rating does not account for power play goals, any changes to a team on the man-advantage can only be in the negative if they give up a short-handed goal, which Hohl notes skews some player types, and does not accurately measure outscoring. Instead, Hohl points to 5v5 Goal% as a more descriptive statistic and shot metrics as better predictive statistics.screen-shot-2016-10-25-at-1-50-58-pm

. . . . .

Even with other organizations delving into analytics and growing evidence that it can give a new perspective on player and team evaluation, there are still many organizations that solely study the traditional approach. John Tortorella, head coach of the Columbus Blue Jackets, has clearly stated his opinion on analytics: “I’d rather spend time doing that than listening to this crap about the Corsi and the Fenwick, because those stats do not apply. I don’t care what anybody says in this league, they don’t apply to the game of hockey”

Jim Benning, General Manager of the Vancouver Canucks, also sticks to traditional analysis because of his experience with the Bruins in 2011: “We won a Stanley Cup in Boston and we didn’t use analytics.” While the Bruins may not have been built up using an analytical approach, they were not coincidently a successful team by those metrics, because analytics can quantify what the eye-test shows to be successful. Rather than discrediting analytics, perhaps a General Manager like Benning should consider supplementary analysis since the eye-test analysis he relies upon has not made the Canucks any better.

Wilke pointed to the Canucks as a team that “really needs to focus on relative metrics right now.” By studying the data, rather than just the eye-test, the Canucks could “pinpoint the guys who are hurting the team the most and then go back to the video to see if there are specific behaviors (pinching too deep, backing off at the blue line, etc.), that are causing those poor possession measures.”

Los Angeles Kings president and general manager Dean Lombardi has also expressed doubts about integrating analytics. “What do you think my chances are if I go downstairs and give Darryl Sutter a computer sheet and say, ‘Why don’t you use this?’” Lombardi continued, “the one thing that is missing in all of this is no one has shown the correlation with winning.”

Stimson responded to Lombardi’s comments on the lack of correlation, saying that, “unless you’ve been asleep for the last five to ten years, you know shot metrics correlate to winning.” Iyer echoed Stimson, explaining how many analytics dismissals stem from a lack of understanding: “Dean Lombardi’s statement that there isn’t a correlation between winning and analytics is not entirely correct as we’ve been able to identify measures that are predictive of team success and as such, have been able to create models that predict how well teams will perform in the regular season with decent success.”

“Oftentimes when an industry goes through a massive change, like the adoption of analytics as a guideline, there’s a bunch of ‘unlearning’––discarding outdated ideas that used to be held as gospel,” which is needed in order to move forward, Wilke said.

. . . . .

Even if more teams integrate analytics, there will always be discussions on how exactly to balance the perspectives, Hohl said. “For the eye-test, it will always be difficult to balance the domain expertise with quantitative methods. Really, the best process is to constantly re-evaluate your process. Look back at how your team’s experts are performing and why/where/how they perform well or poorly. Study where one’s biases lie. Then adjust.”

General Managers must always consider how to best help their team, which in many cases may include analytics. Sznajder noted that “Skepticism is always good, but you should also be skeptical of your previous beliefs and try to take in as much information as you can. It is impossible for us to watch every game and know everything about every player in the league, so getting as much information about their performance on the ice helps.” He continued: “Yes, having a good culture is important and there are some things you can’t measure, but at the end of the day you want to assemble the best group of players you possibly can.”

. . . . .

“Analytics in hockey is still very much in its infancy,” Iyer said, though he added that “the field is rapidly expanding, making it easy to find sensational work.” Wilke, Stimson, Sznajder, and Hohl all point to micro-statistics as the next wave of analytical exploration. “Micro-statistic tracking of things like zone entries, zone exits, passes, forechecking can help us not only learn why some teams and players are statistically better performers than others, but also maybe how to improve our output metrics, optimize usage, and improve player development,” Hohl explained.

Sznajder acknowledges that there are still challenges facing the field, like “most of the data [being] kept proprietary and a large chunk of the data is mostly just trivia.” Having consistent, streamlined data would assist the process by creating reliable, uniform metrics. This would enable teams to study data and understand what they are looking at, because “teams might be just making assumptions about what they think is important when they’re presented with these stats and could be misled because they don’t know what to look for, other than what they believe is important.”

The evolution of these traditional mindsets to include analytics is an ongoing process. Although the 2015-2016 Pittsburgh Penguins won the Stanley Cup, a team that integrated analytics and excelled in all areas, “the Penguins still have Sidney Crosby, Evgeni Malkin, and Phil Kessel. Crosby is still the best player in the league, so it’s easy to dismiss the small, but important, [analytically significant] gains made by doing things like trading Scuderi for Daley,” Wilke says.

To expedite the process, the NHL may need a “moneyball” moment, Wilke suggested. “Until one of those small market teams that obviously uses data to help build their rosters wins the Cup, you’re going to continue to see this kind of pushback.”

← Back to SportTechie