This piece began as an idea to highlight the deficiency in contextual statistics to quantify a true defensive impact at the team level. What exactly is good team defense if all we do is use shot metrics against to determine defensive ability.
Then I went to one of my favorite Tableau visualizations, prepped and maintained by the tireless efforts of Corey Sznajder for some more insight into my preconceived notion that we don’t measure defense properly.
Tracking for this effort is manually coordinated and Corey is a machine, able to track a significant number of games, but still paling in comparison to the untracked games that spill over into the off season for a retrospective rather than up to the minute analysis. Times are indeed difficult for a multitude of people, but if you have the ability to support some truly innovative and important hockey data work, please visit Corey’s Patreon page.
The public sphere sorely misses the granularity attributable to Corey’s data, providing clear context. In absence of this data, the focus naturally falls to shot metrics to determine overall defense. After all, even with some ambiguity, it’s more beneficial to have a measurement of something, rather than guessing at nothing. Currently, Corey has tracked these games in 2020-21, with Philadelphia leading the way, and Carolina lagging by a handful of games – no surprise considering Sznajder is a Hurricanes fan.
There are a multitude of pages and a lot to unpack on the Tableau page, so if you are interested in seeing diving a little deeper than scraped data from NHL play by play sheets via Evolving Hockey or Natural Stat Trick, you can quickly fall down a rabbit hole.
Editor’s Note: Drafting is only half the battle! Get an edge on your competition with our Season Tools that are packed with rankings, projections, a trade evaluator, start/sit tools and much more. And don't forget to use promo code PUCK10 to get 10% off. Click here to learn more!
WHAT IS DEFENSE? My Microstats Love Story
Modern efforts to measuring defensive play lacks nuance characterized by the inclusion of microstats for greater detail, applying a more accurate picture of team defensive structure. Most defensive play is analyzed via goals, shots, shot attempts, and with some context adding zone entry efficiency and hits. That entire dataset is classified under shot and goal suppression, and I appreciate isolation from the overall concept of team defense. Most of the time they are the end result of a chaotic sequence leading to a breakdown, or an exploitation to create a scoring chance and in the end measure the effectiveness of tactics. The point of tracking events is to verify their impact on shooting ability and on-ice effectiveness.
However, shot and goal suppression aren’t true measures of overall defense, and why microstats or context matters. Corey’s data focuses on specific events, but for a more general manner of measuring overall defense, counting changes in possession would provide some much-needed context.
Tracking these events could be as simple as a tick in a team column, or as complicated as event time in order to detect rates or performing time based analysis, and then making comparisons leading to insights on influencing shots and scoring chances.
When the team doesn’t have the puck, they are playing defense.
Defense is the structure in forcing transitional play, when not in possession of the puck to retrieving it. That philosophy could also entail a negative trait such as allowing shot on goal with a subsequent change of possession – indicative of allowing a low danger, lengthy shot that a goaltender shouldn’t miss for example – turnovers, forced turnovers, boards play, puck dump ins without retrieval, among other methods of possession changing hands.
Individual aspects of the forecheck can be measured and compared among established norms as well when sculpting systems at different strength. Player and team level cards can be created and used as baselines. We’ve seen the iPads on the benches, that’s the sophisticated end result that the public sphere misses on an up to the minute basis (noted time lags for publishing).
Offensive forechecking, despite the title and location on the ice, is a defensive scheme, because the team with possession has the puck. To get the puck back, parallel skills are required to those utilized in the defensive zone without the puck, without the same intensity to absorb risk of getting caught unprepared heading back the other way and flowing within team structure.
Good teams will generate chances off the rush, retrieve the puck for multiple opportunities in an individual zone entry, by a heavy rebound presence to force secondary shots from net front, or using heavy forechecking tactics to get possession of the puck back and begin the cycle anew.
A general look at teams that do this, can be found on this Viz. According to the tracked games, the average individual scoring chances per 60 are 3.05 and the average shots of a rebounds per 60 is 0.366.
Editor’s Note: Don’t forget to download the NBC Sports EDGE app to receive real-time player news and updates. Plus, it allows you to easily track your favorite players. Get it here!
I made the axes disproportionate in the above to widen the gap for clarity, visit the Tableau page for the range of players.
Florida and Colorado are capitalizing on Average Rebounds/60, while Detroit and Arizona (both with low games tracked), are near the lower end of the graph. Interesting to associate these leaders in rebounds while being near the top of the list in generating shots of the rush/60, while excelling at the cycle game as well. The transparent circle in the middle of the chart signifies league averages.
A comparison of teams shot generation between rush and forechecking chances is shown below, with Colorado and Florida highlighted. These metrics are rate stats on a per 60 basis, normalizing the lack of tracked games uniformity. Colorado and Florida lead in the rate of rush shots per 60 and have a high rate of generation off the cycle – and absolutely deadly combination.
An effective forecheck is an important defensive concept and was also featured in a presentation at the Ottawa Hockey Analytics Conference by Stathletes, Meghan Chayka to illustrate score effects.
Similarly, stopping teams and forcing turnovers in the neutral zone isn’t always reflected in zone entry attempts and can remain unrecorded as a significant event. A good NZ forecheck is the precursor to zone entry attempts, so maybe there’s more to the datasets involving zone entry attempts that can provide more granularity. Perhaps there are so few possession changes in the neutral zone to adversely affect results and overall perceptions; but in the absence of that dataset, it’s difficult to gauge importance and efficiency.
A striking piece of perception is the Leafs being effective at setting up after a zone entry considering they have struggled incredibly to gain entry with the man advantage. Yes, they’re deadly when in zone, but this graph paints a striking difference between the perception while watching, and what the underlying data reveals.
I wanted to include this viz, because of the the labeling in each quadrant that, well, kind of says it all. The chart is a comparison of the scoring chance generation efficiency on carry-in zone entries. Teams in the upper right quadrant are defined more as a rush team, while in the lower left quadrant don't generate a lot off the rush - and effectively waste a lot of opportunities.
Interesting to see the Blues picking their spots. Also of interest the New Jersey Devils, generating a 55% efficiency at gaining the zone, but only getting chances 23% of the time, just slightly below league average.
Utilizing the microstats to provide more context can influence a lot of the systems and structure that teams utilize in game. With real-time stats available, implementing changes on the fly, in-game opens up a world of new possibilities for coaches and players.