A few weeks ago the NHL finally unleashed its long-awaited player tracking data onto the public. Now it’s time to dissect what it all means.
Sometimes, all it means is “that’s cool!” — and that’s OK. Sometimes a fun little tidbit is all a stat has to be. But it’s still worth looking into how meaningful all the new numbers are and what bucket they fall into: relevant info or interesting trivia.
Advertisement
Each new stat poses its own interesting questions worth answering and that’s the goal of this series: diving into the new data to see how much it matters.
Let’s talk about the big one: zone time.
Is zone time repeatable?
The first question is always the same: how reliable is this new metric? It’s important to know whether it’s measuring a consistent talent that shows stability. Obviously there will be some variation year-over-year, so at the very least, we want to see a decent signal within that variation.
Using R2, which measures the variance from the dependent variable that can be explained by the independent variable, we can see how well correlated each stat is to its previous season. An R2 of one would be a perfect relationship, an R2 of zero would be a non-existent one. The closer it is to one, the stronger the relationship.
Here’s how repeatable each zone time stat is. Zone time percentage is the percentage spent in the offensive zone compared to the defensive zone (i.e., with the neutral zone excluded).
Even strength
Offensive zone: 0.63
Neutral zone: 0.33
Defensive zone: 0.59
Zone time percentage: 0.61
Special teams
Power play offensive zone: 0.31
Penalty kill defensive zone: 0.53
While the neutral zone is important to control, how much time a team spends there varies a lot more from season to season than either the offensive or defensive zone. Both of which have strong repeatability.
For special teams, there’s a strong relationship in keeping teams out of the zone while short-handed, but surprisingly a much weaker one on the power play. There’s still a faint bit of repeatability there — just not as much as the other zone time scenarios.
How does zone time compare to other metrics we already use in terms of repeatability?
Now that we know how repeatable zone time is, it’s important to compare that to what’s already available. The reason on-ice shot metrics became so popular was partly because their larger sample size (relative to goals) made them more stable from year to year — and within the season. The signal was faster to accumulate, more reliable and at one time more predictive. It’s worth seeing how zone time stacks up.
Advertisement
We only have one season of year-to-year repeatability for zone time, so that’s all we can compare to for shots, shot attempts, expected goals and goals. For further context, I’ll also add PDO (shooting and save percentage).
There’s going to be some variability from season to season where one matters more than others, especially as the game evolves. That means it’s important to keep checking and re-checking what matters most between zone time and the other on-ice metrics.
In terms of year-to-year repeatability, here’s how stable each of the metrics compares in reliability listed in order. (Because zone time isn’t score-adjusted, I didn’t use score-adjusted versions for any of the stats below.)
Even strength
Zone time percentage: 0.61
Shot attempts: 0.57
Shots: 0.56
Expected goals: 0.55
Goals: 0.33
PDO: 0.03
Power play
Expected goals: 0.47
Shot attempts: 0.37
Offensive zone time: 0.31
Shots: 0.21
Goals: 0.15
Shooting percentage: 0.002
Penalty kill
Defensive zone time: 0.53
Shot attempts: 0.48
Shots: 0.34
Expected goals: 0.32
Goals: 0.22
Save percentage: 0.11
It should be clear from above why analysts spend a lot of time focusing on shot rates and expected goal rates: They’re a lot more reliable than goals (and shooting and/or save percentage).
Interestingly, zone time is even more reliable at even strength and on the penalty kill. But not on the power play. I would surmise that in that instance, the biggest thing a short-handed team can control is defending entry into the zone, while the power play dictates the quality of chances.
How well does zone time correlate with success?
Do good teams control the zone more? Yes, of course. It’s not perfect, but it’s a strong indicator.
Using R2, here’s how zone time relates to even strength goal percentage, power-play goals per 60 and penalty kill goals against per 60.
Advertisement
Even strength: 0.53
Power play: 0.20
Penalty kill: 0.30
Unsurprisingly it correlates a lot more at even strength, but there’s still an OK signal on special teams, especially on the penalty kill.
Is zone time more meaningful than other on-ice metrics?
It’s not enough that zone time correlates strongly with goal differential. If we wanted to know how good a team was … we could use goal differential.
Where these stats gain meaning is in their predictive power. What predicts future goal differential best?
The biggest reason shot attempt percentage became such a focal point of hockey analysis was that it was often a harbinger of what was to come. Regression to the mean — where the mean was often closely aligned with who won the shot battle. There are obviously exceptions to the rule and we’ve advanced well beyond “this team sucks, look at their Corsi, bro.” But it’s still an important test to see how well zone time predicts a team’s future ability. And it’s important to put that within the context of the other stats.
So here’s the test: How well does each metric predict a team’s even strength goal percentage, power-play goals per 60, and penalty kill goals against per 60 in the following season?
Again, it’s important to stress that we only have two years of data and 32 data points. It’ll take much more than that to figure out the truth (in-season testing — i.e., how do the first 20 games predict the next 62, and so on — would be more helpful here, but unfortunately is currently impossible). It might be more prudent to do this at the player level, but that’s easier said than done with the way NHL Edge currently operates showing one player or team at a time.
With that caveat out of the way, here are the results.
Even strength
Shots: 0.47
Expected goals: 0.47
Shot attempts: 0.41
Zone time percentage: 0.38
Goals: 0.33
PDO: 0.03
Advertisement
Power play
Expected goals: 0.35
Shots: 0.27
Shot attempts: 0.21
Goals: 0.15
Offensive zone time: 0.14
Shooting percentage: 0.03
Penalty kill
Expected goals: 0.35
Shot attempts: 0.26
Shots: 0.26
Goals: 0.22
Defensive zone time: 0.15
Save percentage: 0.05
That’s a nice reminder about why we use expected goals and other shot metrics over goals and shooting percentages — it’s usually more predictive. There are better ways to measure that, but we’re unfortunately at the mercy of the zone time stats we have. Despite its strong repeatability, zone time does not look to be more meaningful than what is already available. At least not when it comes to 2021-22 data predicting 2022-23 data.
How does zone time correlate with shot attempt percentage?
One of the original reasons that shot attempt percentage gained such prominence was that puck possession was all the rage. We didn’t have data for that, but it was theorized that shot attempt percentage served as a valuable proxy for possession (with some folks going out of their way to track zone time back in the day to compare). That indeed looks to be the case as there’s a very strong relationship between zone time and shot attempt percentage at even strength over the last two seasons (R2 of 0.76).
That’s obviously not a perfect relationship and that’s OK as what we saw above was that meaningful possession is most important. It’s not about having the puck more, it’s about what you do with it.
There’s a lot of room for how well a team uses their zone time — and that’s the next topic of exploration.
Is zone time more important than zone efficiency?
How do we currently measure how good a team is offensively? By looking at how many goals or expected goals they earn per 60 minutes. But you can’t score from the defensive zone. Should those minutes count toward how efficient an offense is? Enter: “zone efficiency.”
Thanks to zone time we can now surmise how much time a team spends in each zone and can then use that time instead of all the minutes to measure how effective an offense is in-zone. Instead of goals per 60, it’s goals per 60 offensive minutes. That’s the idea of zone efficiency.
Here’s what that looked like last year at even strength using expected goals per 60 offensive and defensive zone minutes.
The question that arises from that is obvious: What’s more important, playing in the offensive zone more or making the most of your time in the offensive zone? (Likewise for defense and special teams.)
Advertisement
To measure that, we’re going to the same tests as above, except we’re comparing two variables to a target at the same time using a multiple linear regression.
The target is goals and the variables are zone time and zone efficiency — with zone time converted to an expected goal rate to be on the same playing field. With a multiple linear regression, the more important variable will have a heavier weight (coefficient) in the resulting regression equation.
First, how they correlate in-sample. How much does zone-time and zone efficiency describe goal differential relative to the other?
Even strength
Offense
Zone time: 0.24
Zone Efficiency: 0.17
Defense
Zone time: 0.15
Zone efficiency: 0.26
Both
Zone time: 0.49
Zone efficiency: 0.51
Over the two-season sample, it seems as if both are equally important at even strength — but not when looking at offense or defense separately. There, it seems that zone time is more relevant to offense and zone efficiency is more relevant to defense. My guess is that it’s possible earning zone time is something dictated by players with the puck, but turning that time into something is dictated by the defense.
On special teams it’s a different story: Zone efficiency is what matters most.
Special teams
Power play
Zone time: 0.10
Zone efficiency: 0.47
Penalty kill
Zone time: 0.13
Zone efficiency: 0.44
As was the case before, it’s also important to measure how meaningful each measure is out-of-sample. How well does zone time or zone efficiency explain future goal differential relative to the other?
Even strength
Offense
Zone time: 0.35
Zone Efficiency: 0.06
Defense
Zone time: 0.34
Zone efficiency: 0.07
Both
Zone time: 0.63
Zone efficiency: 0.37
Despite zone time and efficiency both being relevant for different reasons when describing what happened at even strength, it’s zone time that shows to be more meaningful when it comes to predicting the following season.
That’s not the case on special teams though where in-zone execution is everything. If a team has strong special teams because of zone time, but is weak at executing in zone, chances are they could have a bad time the following season.
Advertisement
Special teams
Power play
Zone time: 0.08
Zone efficiency: 0.51
Penalty kill
Zone time: 0.00
Zone efficiency: 0.61
Again, I want to stress that this is a small sample. But for now, it’s all we have and it’s our best guess. While both zone time and efficiency matter, it’s zone time that matters more at even strength and zone efficiency on special teams.
Is zone time more reliable than zone efficiency?
So we know that, for now, zone time is more important than efficiency at even strength and that’s partly because of repeatability. Recall that both offensive and defensive zone time was very repeatable with an R2 of around 0.6. That’s not the case for zone efficiency where offensive efficiency is only half as repeatable (R2 of 0.35) and defensive efficiency doesn’t look very repeatable at all relative to the strength of other on-ice metrics (R2 of 0.09).
That’s also the case on the penalty kill (R2 of 0.14), but not on the power play (R2 of 0.44) which is stronger than zone time. That seems to be the one place where zone efficiency holds a lot of worthwhile utility.
There is some value in it elsewhere, but more so as a descriptive tool to figure out why a team’s offense or defense is struggling. There’s more predictive power elsewhere.
What translates better in the playoffs: zone time or zone efficiency?
How meaningful and reliable these new stats are important for regular season team improvement, but it’s also wise to check in on how that applies to the playoffs as well. Are zone time and efficiency just as meaningful and reliable when comparing the regular season to the playoffs?
In terms of repeatability, the ability to earn more zone time is the biggest thing in a team’s control going into the postseason — though it is close on the power play.
Even strength
Zone time: 0.50
Zone efficiency: 0.34
Advertisement
Power play
Zone time: 0.25
Zone efficiency: 0.21
Penalty kill
Zone time: 0.20
Zone efficiency: 0.04
In terms of describing what happened, it’s zone time in a landslide at even strength (a weight of 0.85 for zone time and 0.12 for zone efficiency). On the power play and penalty kill it’s zone efficiency with zone time having no relationship. Considering the lack of repeatability for penalty kill zone efficiency from the regular season, it makes it difficult to trust which penalty kills will actually execute in-zone defense come playoff time.
Lastly, we look at how each regular season stat was able to predict playoff goal differential relative to the other. This should add some context to the regular season test above as there are twice the data points and the out-of-sample test is in-season (meaning the rosters are more similar). Though there is obviously the caveat of opponent quality and small samples working against us too.
Even strength
Zone time: 0.30
Zone efficiency: 0.23
Power play
Zone time: 0.53
Zone efficiency: 0.04
Penalty kill
Zone time: 0.89
Zone efficiency: -0.20
Here, zone efficiency carries a bit more weight at even strength. But on special teams, zone efficiency looks to be the key.
Do teams with stronger zone efficiencies score or defend better than expected?
One last thing to check in on with zone efficiency.
I wanted to look at whether teams with a high zone efficiency were able to score more than expected (or allow less than expected). Why teams score or allow less than expected beyond “puck luck” is one of the most pivotal questions in hockey analytics. So it’s always worth checking if any new stat can increase our knowledge of what to expect.
Interestingly, there appears to be a negative relationship (R2 of 0.19) between even-strength offensive efficiency and goals scored above expected.
There’s probably a lot more to study here, but this would suggest that teams that take fewer expected goals per offensive minute might be doing so for a reason: patience in finding the perfect shot.
That could also just be a coincidence — there wasn’t much of any relationship when looking at it defensively or on special teams. The relationship also weakens when using shot attempts to measure zone efficiency rather than expected goals.
For those who wanted to play with the NHL data I collected further, I’ve included it in a CSV here.
Advertisement
Data via NHL Edge
GO DEEPER NHL goalie tandem power rankings: Which teams are best set up for success?
(Photo of Cale Makar and Sidney Crosby: Jeanine Leech / Icon Sportswire)
