My first article focused entirely on save percentages. I thought it would be helpful to have a look at the other side of the coin – shots – and how the two interact.
It’s pretty well known that Sv% has risen and goals per game have dropped over the years, but what is the long-term trend in shots per game? And do the playoffs see a drop off in shots from the regular season?
Here’s the league-wide average for shots per 60 minutes in the regular season and the playoffs since 1983-84:
You can see that there are slightly fewer shots in the playoffs most years, but that’s less true recently. As always, the playoffs are more volatile than the regular season, because fewer games are played. I’d interpret this graph as showing:
- Pretty stable Shots/60 from 84-97 and 06-16.
- A low-shooting era from 98 to 04. I assume the uptick in 06 was related to the post-lockout crackdown on obstruction. It’s interesting that the drop in Shots/60 from 97-98 is even larger than the rise from 04-06.
What’s most striking to me is that post-lockout, Shots/60 has returned to nearly 80’s levels. Note that Shots/60 over the past 3 regular seasons is essentially identical to 1987, but (as we see below) scoring is down by a full goal per 60 – almost 30%. The difference in goal scoring between the 80’s and the present is almost entirely due to shooting percentage.
Here’s the graph for Goals/60:
And here’s shooting percentage (the inverse of the Sv% graph from last time – it’s helpful if all the lines slope in the same direction):
You can see there’s a strong resemblance between the Goal/60 and Sh% graphs. Here’s the relationship between Goals/60 and Shots/60, then between Goals/60 and Sh%:
I was also curious to see if there was any relationship between Sh% and Shots/60.
The graph shows a weak positive relationship for the Regular Season, and a very weak one for the Playoffs. It looks like (at least in the Regular Season) years with higher average shooting% also tend to have higher average shots/60. But the graph is misleading, because of the trends noted above: a general decline in shooting%, and a low shots/60 period in the 90s. It’s not really fair to bundle different eras together. We can see this clearly by labeling the data points chronologically:
I’ve separated out the three eras I identified above. You can see that the early years had more shots and a higher (but declining) Sh%. The pre-lockout years had few shots and slightly-declining Sh%. Post-lockout, shots/60 returned to early-90’s levels, but Sh%, after an initial surge, continues to decline.
I also included G/60 lines, showing how the two variables contribute to scoring. Again, it’s clear that goal rates are impacted much more by shooting percentage than shot rates – thus the G/60 slopes are quite low. Another way of looking at this is that league-wide Shots/60 has been quite stable compared to Sh%.
To make things more concrete: an increase of 1 shot/60 next year would have about the same effect on scoring as an increase of 0.3 in Sh% (-.003 in Sv%). Either will amount to an extra seven goals per team per season.
Here’s the same chart for the playoffs. Again, there’s more variability than in the regular season, but the three eras are still quite distinct:
By the way, the Goals/60 lines in the charts above are not quite straight – they do curve up as shots/60 approaches zero, and tighten as Shots/60 grows. Here’s a chart showing G/60 at extremes of Shots/60 and Sh%:
The green box is the area covered by the Playoff Sh% and Shots/60 chart.
Returning to the question of how Sh% and shot volume are related on a league-wide level, I compared each year’s Sh% and Shots/60 to the average of the surrounding four years, which should provide a better comparison across eras:
There’s a decent positive correlation for the regular season – years in which shots/60 were high relative to the surrounding years also tended to see higher Sh% than the surrounding years. In the playoffs we don’t see much of a correlation – if anything, there’s a weak negative relationship.1
Here are my takeaways:
1. The league-wide scoring rate is much more influenced by shooting percentage than by shots per game. The average shots per game seldom varies enough from season to season to be meaningful.
2. There was a significantly lower-shooting era which began abruptly (and mysteriously) in 98 and ended with the 05 lockout.
3. There’s some positive relationship between league-wide shot volume and Sh% in the regular season, but not in the playoffs.
We can add the 98 shot plunge to the list of peculiarities I’d like to explain. (A list of unanswered questions concludes my previous post. I’ve poked around, but haven’t made much headway in answering them.)
I suspect that the long-term decline in Sh% is predominantly due to improvements in goalie equipment, skill and tactics, rather than changes in the defensive skill and tactics of skaters. I reason that if improvements in team defensive since the 80s had significantly outpaced improvements in team offence, we’d likely see a consistent decline in Shots/60, not just in Sh%. The best way to test this theory this would be to apply some form of shot quality metric to historical data. While there have been a few attempts to measure shot quality, to apply them historically would be prohibitively time-consuming.
Update Apr 10, 2018:
An article by Woodguy tipped me off to another factor I overlooked: a significant decline in powerplay opportunities per game. I may explore this further in the future.
A lot has changed in the NHL since the 80s, but by far the biggest change has been the steady improvement in goalie performance. In 83-84, Billy Smith led the league with a .896 Sv% (min. 40 games). Last year, Mike Condon finished dead last with .903.
Now that we’ve got a handle on league-wide trends in shots and Sv%, I’d like to take a closer look at how they impact individual teams, and individual playoff series.
1 I also tried comparing each year to just the previous season, or to the average of the previous 3 seasons – the R2 were .34 (RS) and .05 (P) for the last year, and .28 (RS) and .02 (P) for the last 3 years.