NHL Edge Data Deeper Dive: Team Burst Rates vs Team Shot Rates

Does playing faster mean playing faster?

Continuing my look into the NHL’s Edge data, I want to go a bit deeper into a few topics, as well as keep the topic of each article more focused. With that in mind, I wanted to examine the relationship between hustle (how often teams exceed certain speed thresholds) and pace (the rate of shot attempts). Let’s see if there’s anything to support coaches’ love of hustle.

Methodology

My methodology for this article was straightforward. I took the bursts per 60 minutes I calculated in my previous article and compared it to corsi and expected goals for each team per 60 minutes using Natural Stat Trick’s all situations numbers for the 2021-22 and 2022-23 seasons because that is the format burst numbers are available in.

For completeness, I looked at 18+, 20+, and 22+ MPH cutoffs. However, I suspect that 22+ MPH is not a very useful metric, due to how infrequently that speed is reached by all but a handful of players. On the other hand, every NHL player with a recorded top speed (four players do not have a top speed recorded) can exceed 18 mph and about 75% of the league reached 20+ mph 10 times or more. This tells me that 18+ mph is a useful proxy for skating hard as it is attainable for most players without going all out.

I chose corsi to measure the relationship between hustle and shot quantity and expected goals to measure total shot quality. In addition to per-hour rates, I also looked at the relationship between bursts per 60 and expected goals per unblocked shot attempt (fenwick) to see if teams that hustle more create better chances rather than winning the expected goals battle through volume of chances. Lastly, I repeated the exercise on the defensive side, comparing burst rate to corsi and expected goals against rates and expected goals per fenwick against to see if hustle also had a defensive impact.

Results

Corsi Rates

The relationship between hustle and corsi for per 60 is quite strong. In fact, we can say with 90% confidence that every speed cutoff for burst rate is positively correlated with CF/60, though the relationship weakens as the cutoff gets higher. I suspect this is due to the aforementioned infrequency of bursts at higher cutoffs. The relationship gets even weaker when the four fastest teams are removed.

On the defensive side, the pattern is repeated, although the correlation is negative. This means the more hustle-y teams are effective at suppressing shot attempts against, though only the lowest speed cutoff meets a 90% confidence rating.

In sum, the frequency with which a team has bursts of hard skating does seem to have a relationship to both shot quantity generation and shot quantity suppression.

Expected Goals

Here we see a similar pattern to the relationship between hustle and corsi, though the correlation is not as strong as the correlation with corsi. On the offensive side, we can say with 90% confidence that both the 18 MPH and 20 MPH cutoffs are positively correlated. On the defensive side, we can once again note that the correlations are overall weaker than on the offensive side, and note that we can only be more than 90% confident in the correlation with the 18 MPH cutoff.

An open question on this relationship is how much of the expected goal rates are driven by overall volume (piling up low xG shot attempts vs getting high-quality chances). On the defensive side, there is also a question of how much hustle results in simply holding the puck more and how much hustle is related to defense without the puck.

Expected Goals Per Fenwick

Lastly, we examine the relationship between hustle and shot quality. The first thing that stands out immediately is that none of the correlations are strong enough to say with 90% confidence that the correlation is not zero. In the results available, there is an interesting reversal of the trend on offense, with the highest speed cutoff having the strongest relationship and the lowest speed cutoff having the weakest. Overall, however, none of these relationships are strong enough to be considered significant findings. Looking at the four fastest teams (Colorado Avalance and Edmonton Oilers for both seasons) reveals a hint as to why. Except for the 2022-23 Oilers, they all fall middle of the pack in terms of expected goals per fenwick. More seasons of tracking data may shed additional light on the value of hustle in increasing shot quality.

On the defensive side, we observe a minor negative correlation between team hustle and average shot quality against. There is very little difference in the correlations between each speed cutoff and none are strong enough to state with 90% confidence that there is a relationship.

Conclusion

Hustle (defined as bursts over 18 MPH per 60 minutes) does appear to have a decently strong relationship with offensive creation. Teams that have more frequent high-speed bursts also tend to have higher shot quantity rates. However, that relationship does not appear to extend to shot quality. On the defensive side, hustle is correlated with decreased shot quantity allowed, though the relationship is not as strong as on offense. Similarly, there is not a significant relationship with shot quality suppression. Overall, increased hustle does appear to correlate with positive team outcomes.