Hey folks,
I did a study on NHL scoring levels since the 2004-05 lockout, inspired by the recent discourse on reduced goal scoring and the alternatives being discussed. I posted the work on my here a while ago, but now I'm transcribing it and sharing with you guys.
Thoughts? Comments?
I did a study on NHL scoring levels since the 2004-05 lockout, inspired by the recent discourse on reduced goal scoring and the alternatives being discussed. I posted the work on my here a while ago, but now I'm transcribing it and sharing with you guys.
There is no question that declining scoring levels in the NHL has fans, players and executives concerned alike. Many believe the league is trending towards a return to the “Dead Puck Era”, a period spanning from the late 1990’s and into early 2000’s marked by relatively low scoring levels. Following the 2004-05 lockout1, new measures such as the introduction of the trapezoid, the two-line pass, the reduction in goalie pad size, the tough stance on “bear-hugging” and the shootout seemed to solve scoring woes as teams scored more than 3 goals for per game in 2005-06. However, scoring levels promptly fell to 2.88 in the following season, beginning a downward trend that has persisted into 2012-13, reaching a post-lockout low of 2.65.
Although General Managers think oversized goaltending gear has something to do with it, factors like defensive-minded coaching, a return of “clutch and grab” and shot blocking are also points of contention. I’m not exactly sure whose judgment I should trust, so I figured I’d take matters in my own hands and come closer to the truth myself.
Let’s start by zooming in on the previous chart and focusing on post-lockout trends.2
At first glance, it seems as if players are having a harder time putting the puck into the net, evidenced by a decreasing shooting percentage. Shots per game have moved less significantly and looks to be a minor factor.
But is it really any harder to score?
Both goals and shooting percentage adds up scoring events in all three situations within hockey – even strength, power play and penalty kill. Intuitively, one should assume that getting more power play time results in more goals. After all, it should be easier to score when there is one, two or even three less opponents to play against..
Interestingly, power play conversion rates haven’t meaningfully changed since the lockout. What has changed – and quite dramatically – are the numbers of power play opportunities teams get on average. Referees called 43% less penalties in 2012-13 than they did in 2005-06. Accordingly, there were 41% fewer power play goals in 2012-13 than in 2005-06. With so fewer power plays going around, we should expect teams to score less on average. By how much?
Indeed, as power play goals per game decrease so does overall scoring.
Linear Regression, with PPG/G per season as input and GF/G per season as output
Linear Regression, with PPOPT/G per season as input and GF/G per season as output
With an R2 of .9125. In other words, 91.25% of changes in goal scoring can very likely be explained by changes in power play goal scoring. The relationship with power play opportunities is slightly weaker, but nevertheless very strong with an R2 of .8765. It’s very easy to justify the result – power play yields a higher shooting percentage than even strength play. Therefore, the less power play opportunities there are, the less players get to play in higher percentage scenarios and thus, scoring goes down.
Note: The relationship between PPOPT and Goals is more important than the relationship between PPG and Goals. Changes in ES and SH scoring plus PP conversion rates can serve as lurking variables in the PPG regression. Regressing PP opportunities establish a more straight forward relationship between penalties being called and goals being scored.
Intuition can be proven by comparing even strength and power play shooting percentage. The results should surprise no one:
PP shooting percentages seem to have trended down since the lockout, implying less scoring on the power play. However, we have already seen that conversion rates are not meaningfully different, perhaps hinting at teams electing to get more shots through instead of setting up a higher percentage chance on the PP. There are significant changes in ES shooting percentages, which will be discussed later. First, let’s look at the ES scoring pace:
Teams have actually been scoring more in even strength since the lockout. However, this information does not mean players have had a better time scoring during ES play. We have already established power play opportunities have dramatically declined since 2005-06. We have also reviewed ES shooting percentages and concluded they haven’t changed much. The answer therefore, very likely lies in ES shooting rates.
Just like that, it seems as if changes in ES shooting rates go hand in hand with ES goal scoring rates. There’s more even-strength play, so more ES shots are made and more ES goals are scored. To confirm this theory, let’s look at the change in PP TOI allocation since the lockout.
There’s a 43% reduction in PP TOI, which is the same percentage as the change in PPOPT (intuitively they should be almost perfectly correlated). In order to calculate ES TOI, the total ice-time per season is subtracted by the PP TOI, times two (since PP TOI = SH TOI).
To slightly ease calculation, TOI was formatted into decimals instead of conventional time measurements.
ESTOI seems to go along well with ESG, but to what extent?
There is a strong positive relationship between ESTOI/G and ESG/G (R2 = .6717), meaning most of the changes in ES goal-scoring can be explained by changes in the amount of ES time players get, proving the intuition established earlier.
Still, one would expect a stronger correlation if we were to assume scoring hasn’t gotten any harder. As seen earlier, ES shooting percentages are trending down since 2007-08, perhaps more so if data from 2005 and 2006 were aggregated. ES shots per 60 are also on their way down:2
While ES SV% has increased since the lockout, as expected given shooting percentages have gone down.
These numbers could serve either as proof that goaltenders are better at stopping the puck or that defensive systems are stifling offense, or a combination of both. They also explain why ESTOI and ESG aren’t very strongly correlated. Regardless, it’s quite clear players are having a hard time getting quality shots through in even strength play. But as seen before, converting in the power play doesn’t seem to be any more difficult. Why would that be?
Some have said blocking shots makes scoring more difficult, but there are studies that have shown blocks have little to no impact in terms of winning, or even helping their goaltenders out. There is also no relationship between save percentage and shot quality, meaning allowing more or less shots through isn’t going to significantly impact the goaltender’s save percentage. Quantifying the effect of goaltender equipment on scoring is very difficult due to changes in the game that yield more high quality shots and changes in the goaltenders’ techniques themselves. What is left to see whether the argument that “clutch and grab” hockey holds water. As seen already, the number of PP opportunities have declined by over 40% since 2005-06 and it explains changes in goal scoring very well. But is calling less penalties making it harder for players to beat goaltenders during even-strength play?
Intriguingly, changes in PP opportunities can be used to explain changes in ES save percentage extremely well (R2 = .9335) to the same degree it explains changes in overall goal scoring. There are reasons to believe the relationship is misleading – it can be argued the relaxation in refereeing coincided with goaltenders accustoming with post-lockout equipment standards. Still, the regression can also be employed as evidence that more hooks, trips, interference, etc. are going uncalled, making it harder to get pucks through in good spots. PP opportunities can also be used to explain changes in ESS/60 very well (R2 = .9042) which further fuels the “clutch and grab” argument.
Lastly, for better visualization of the impact of PP opportunities, each season was adjusted to have the exact number of power play opportunities, with power play and even strength ice time altered accordingly to reflect the differences in special teams’ time.
Although scoring increases by a fair margin, power play opportunities in isolation do not make up for the differences in scoring between the 2005-06 season and subsequent ones. This should be expected as ES shots per 60 minutes decreased and goaltender SV% increased since then. The adjustment cannot control for the way the game has been refereed. So if “clutch and grab” is indeed responsible for reductions in ES scoring, it can very well explain the differential that still persists.
What to make of all of this? First, the reduction in power play opportunities can be used as a reasonable justification for not only the reduction in scoring, but also for the increase in even strength goaltender save percentage. Less power play time means more of the game is played at even strength, where players shoot at a lower rate. Less power play opportunities is associated with higher save percentages, most likely due to an increase in uncalled obstruction penalties.
Because power play opportunities can largely explain reductions in ES goal scoring and shooting rates, it seems unreasonable to pin the decline in scoring on goaltending equipment. It may be that large chest protectors, long pads or the sheer size of goalies make shot stopping easier, but nowhere to the degree that less power play time does.
Defensive systems may be a reason behind the scoring reduction, but it’s inclusive whether intrinsically better defense is being played, or if systems that encourage more obstruction are thriving under laxer refereeing. The PPOPT regressions suggest the latter.
Thank you for reading my first blog post! Feel free to share your thoughts in the comments section. If you thought this post should be read by more people, don’t hesitate to spread it around.
Cheers!
Endnotes:
1 Every time I use the word “Post-Lockout”, I am strictly referring to NHL season following the 2004-05 lockout, including the lock-out shortened 2012-13 season.
2 Expect a blog post directly comparing Dead Puck Era and Post-Lockout metrics in the near future.
3 I calculated even strength shooting percentages by using data available at behindthenet.ca. ES shot numbers go only as far back as 2007-08, and I was unable to find a publicly available source that had data from earlier seasons. I attempted to aggregate numbers found at NHL.com, but they do not list shots taken while shorthanded or at even strength. They do split saves by on-ice situation so SV% numbers are easily retrievable.
4 These numbers do not include shots that made it to an empty net as they were based off the numbers goaltenders faced. This was done due to circumstances cited in Endnote #3. Adding ENG to the mix will not help because NHL.com does not categorize ENG by situation. Shot s that become an ENG are very infrequent, however, so omitting them would most likely make no difference.
Thoughts? Comments?
Last edited: