A Primer On NHL Analytics And How Experts Use Them (Part 2)


This is part two of this two part series by Shayna Goldman that shares insights from six hockey analytics experts about how they integrate analytics into hockey analysis. Part one explored how analytics can detail information about offensive and defensive production, predict scoring, demonstrate an individual player’s value, how to build a penalty kill, build narratives and provide supporting evidence for those narratives, explain what happened on the ice, and create visual models that can be used in place of charts full of numbers.

. . . . .

Emmanuel Perry built Corsica, a database containing a wide range of hockey statistics, including traditional stats and more advanced analytics. While Corisca is all Perry’s original work, it was modelled after and intended to fill the void left by WAR On Ice.

The data Perry uses for Corsica is sourced from the NHL, Sportsnet, and ESPN. “I capture the play-by-play and shift reports provided by the league to construct an enhanced record of all recorded events that occur over the course of every available NHL game,” Perry explained. He acknowledged that collecting data like this is a starting point for many others in the analytics community. When collecting and studying this data, Perry added that he hopes “only to learn about hockey––the nature of the game itself and how various players and teams perform.”

Granted that the data is “analyzed responsibly,” there is a lot of information that is important from the data collected, including “which metrics are indicators of future playoff success and which are omens of regression? How much does a particular skater contribute to their team’s offensive production? When should a coach pull their goalie for the best chance of tying the game? These are questions that require empirical data to answer with any scientific validity.”

Get The Latest Sports Tech News In Your Inbox!

Perry explained that analytics are the type of evidence that could be used in standard hockey analysis. Often, hockey analysis turns to former NHL players; however, Perry noted that “respectfully, having played NHL games is not a substitute for actual evidence. I think responsible use of analytics prevents dogma from being perpetuated across generations of hockey fans and encourages new ideas.”

For fans looking for statistical evidence, Corsica could be a useful resource. Perry thinks that “anybody who chooses to seek this stuff out has a right to comprehensive and reliable data; I hope Corsica can provide that for at least a few people. I want it to be a launching pad for anyone interested in digging a little bit deeper to learn more about the game we love.”

Perry continued, “I think fans should investigate analytics to the extent of their interest and no further. If analyzing numbers in any way detracts from your enjoyment of hockey, what’s the point?” But he hopes that fans do not outright dismiss what analytics can offer.

For the fans looking to integrate analytics into their hockey consideration, Perry thinks that Corsi is an important statistic to understand. “I’m well aware of the criticism toward it, even within the statistics community. Despite its limitations, Corsi earned the reputation it has as a new-wave metric and a considerable upgrade over traditional measures like plus-minus. Rates of shots for and against are valuable indicators of quality at both the team and skater level – where, in the latter case, I would recommend using relative stats.” And when looking at goaltending statistics, Perry advocates for Goals Saved Above Average (GSAA) as opposed to Goals Allowed on Average or Win totals.

While Perry stressed that only fans that take an interest in analytics should delve deeper, his opinion differed for those who cater to a larger audience. “If your opinion on hockey is relied upon or consumed by a large audience, I believe it is your responsibility to understand the game in enough depth to provide valuable insight. If you make a living talking about hockey, I can’t think of a reason you shouldn’t wield a decent understanding of modern statistics.”

. . . . .

Sean Tierney’s dataviz work, in addition to his articles at Hockey-Graphs and Today’s Slapshot are valuable analytics resources as well. Tierney describes himself as a data scavenger––citing Emmanuel Perry’s Corsica, Ryan Stimson’s passing data, Corey Sznajder’s zone entry and exit data, and Hockey-Reference as some of the myriad of sources he uses.

The data Tierney collects is utilized to create visualizations via tableau. Some of the graphs Tierney creates are interactive, allowing fans to use filters to view different aspects of the charts. “Often, the graphs I make communicate some straightforward ideas that don’t take long to draw an insight from. For example, last season I created game charts for every NHL game from January through to the Stanley Cup Final. The graphs showed each player’s Corsi differential, their time-on-ice, and their individual shots. The graphs gave a quick view of some of the key advanced stats takeaways from every match,” Tierney explained.

The scale in which Tierney does his visualizations and analysis varies based on the phase of the year. For example, during free agency or the NHL Entry Draft, player-specific graphs are more fitting. Other times, though, call for team-level visualizations to “help give a sense of how teams are performing based on a variety of metrics.”

Tierney has analyzed at the player-level through passing data to identify the best passers in the NHL and goes even further by looking at NHL trends like “outliers, team systems, [and] player positions.” That type of analysis suggests which players could be considered elite, or show which players are not effective on a team’s roster.  

Although an individual player may not find any actionable takeaways from Tierney’s visualizations, they may offer insights to a coaching staff––such as finding the best strategy to fully utilize a particular player’s talent. “For example, when graphing Stimson’s passing data, it was possible to see which players were good at creating passes that directly led to shots for their linemates. Using this data, I looked at game film to explore strategies used by the game’s best passers to create passing lanes. A coach might use these data-driven insights to establish a game plan that facilitates passing.”

NHL front offices also could find value in this information by viewing the team-level visualizations, perhaps by glancing at a scoring chances graph to quickly note whether or not players are exceeding or struggling to meet expectations.

In general, Tierney designed the visualizations to accommodate hockey writers and fans. “The idea is to condense lots of information into singular views that allow for quick reference or for use in articles to quickly illustrate a point without a huge table of numbers with decimals.”

The work Tierney has done thus far has significant potential for moving hockey analysis­­­ forward––including a more standard integration. “I think the way forward is for analysts to continue moving from stats and viz as stand-alone information and to give numbers context by connecting data to game video.” Tierney credits the work of Stimson, Prashanth Iyer, and Charlie O’Connor for inspiring him to move in that direction as well. “[Stimson, Iyer, and O’Connor] have done some excellent work fusing together video and hockey stats to generate insights with real-life context,” showing how analytics can both formulate and support narratives with hard evidence––incorporating analytics into a more standard basis for hockey analysis.

. . . . .

Matt Cane, a writer for Hockey-Graphs, is another resource in the analytics community. As a hockey analytics researcher, Cane seeks to “devise new ways to measure the success of players, teams, and strategies.”

Cane gathers data from a variety of websites, including Corsica, HockeyViz, and Puckalytics, when researching a player, team, or game, as well as to reference custom metrics those websites have created. For more specific analyses, Cane refers to his personal database. “I’ll use my own database when I need the detailed event-level information that those sites don’t have, which is really the key when doing research to create new metrics or test strategies.” Cane’s database provides him flexibility when exploring concepts that have yet to be analyzed by others in the analytics community.

According to Cane, there are two predominant contributions analysts can make to a team or fan’s hockey comprehension: describing or predicting the success and failures of a team or player, and analyzing tactics and strategies. “I think the work done to date has been really good at the former and perhaps a bit slower to tackle the latter.” The first contribution, as Cane explains, is the “most natural question you ask as you get started: who’s the best, who will be the best, did my team make a good trade” and so on. But actually exploring the tactics and strategies that make a team effective can add even more value. Cane cites one common example—understanding the right time to pull a goaltender in the game—as an area where having a solid strategy can become a major advantage for a team. “Player evaluation is obviously valuable too, but you get many more opportunities to tweak your strategy than you do to drastically change your lineup.”

Analyzing goaltending with advanced statistics is an area that still needs improved data, Cane says. “As things currently stand we just don’t have enough details on all the things that were happening before a shot occurred, as well as where the goalie was, where the other players were positioned, etc. And even if you had all of that data, there’s still may be a lot of randomness built into the data that will make it tough to draw broad conclusions from it.”

Although there are setbacks in goaltender analysis due to the data limitations, Cane believes there have been improvements in the analysis thus far. Accounting for the danger of shots, based on the shot location, has been critical in understanding which goalies face the toughest shots. This information helps establish which goalies are the best, since those who are most successful at stopping tough, high-danger shots tend to be the best year after year.

Cane recognizes the effort and growth in developing analytical metrics. He notices a gap in translating that into day-to-day analysis for a team, but is encouraged by the progress being made in shifting towards that direction. “When you look at the insights that have been gained from the passing data work done by Ryan Stimson or the zone entry work started by Eric Tulsky and continued by Corey Sznajder, it’s really crazy to think how much more we know about the game because of their efforts to manually track data that no one else has.” From that data, it can be shown, for example, that a “defense is struggling because one defenseman is being regularly targeted by opposing forwards when they enter the zone.” Cane says that explaining the conclusions that way is easier for a coach to understand and act on than simply stating that the team or player has poor possession statistics.

For fans though, analytics serve a different purpose. Cane hopes that fans are “first and foremost entertained or at least come away with new ideas about how the game works from what I write. I know that not everyone will agree with my findings or methods (and they shouldn’t – it’s good to be critical!), but I hope that I can at least communicate why and how I got to the answer I did well enough.”

. . . . .

The hockey analytics community studies all aspects of the game. Interested fans can learn about different analytical approaches through these different perspectives.

For example, fans may want to understand a player’s career production and place in a lineup without looking at a chart of numbers. Micah Blake McCurdy’s visualizations give fans that option. Here, recent New York Islander free agent signing Dennis Seidenberg is quickly summarized through a variety of charts. Fans can glance over this, to become acquainted with his production in case they were not already.

Alternatively, a fan could want to better understand the trade of forward Taylor Hall from the Edmonton Oilers to the New Jersey Devils for defenseman Adam Larsson. In order to understand the caliber player the Devils traded for, Ryan Stimson used passing data to create a comparison radar chart between Taylor Hall and Sidney Crosby. Although Crosby is still superior, the fact that Hall ranks so closely shows just how talented he truly is. Larsson on the other hand, while a solid defenseman, is so starkly inferior to a defenseman of a closer value to Taylor Hall, like Oliver Ekman-Larsson. These charts depict the value of these players in this way since there is not a direct comparison for a forward to a defenseman. Stimson’s charts indicate that this was a lopsided trade.

Maybe a fan is looking to delve deeper into a team’s performance, giving the fan a deeper understanding of the talent on their roster. That fan could look to Sean Tierney’s data visualizations for reference. In looking at the graphs below, a fan could recognize that Rick Nash, a player that is often criticized for his lack of offensive production, takes a high number of shots. It also indicates that a player like Tanner Glass, who some say brings “grit” and “toughness” to a roster, did not make sufficient contributions to the offense. Tierney also provides an analysis of the defense’s offensive production, showing just how much Keith Yandle contributed. Yandle was often critiqued for his offensive play, but as the visualization shows, the Rangers’ defense needed Yandle for his all-around offensive skill because their other defenseman lacked that.

The fact of the matter is that analytics are being integrated into today’s game. Between the Arizona Coyotes’ hiring analytically-minded John Chayka as General Manager, the Florida Panthers’ front-office overhaul to focus on analytics, and the Stanley Cup Champion Pittsburgh Penguins utilizing analytics to build a winning team, it is clear that at least some NHL teams see the use in these enhanced statistics. So, whether or not fans choose to study analytics, it is clear that they have earned their place as a valuable supplement to the eye-test.