Stimson: More on passing data and the shot quality debate

Follow Ryan Stimson on Twitter @RK_Stimp

Recently, there has been a renewed discussion on shot quality. I noticed several people engaging on Twitter discussing the merits of shot quality, which even led to some recent posts by David Johnson of and You can read those pieces, and I recommend you do, over at

Shot quality exists, folks. What needs to change is how the debate around it is framed. It is not simply how many scoring chances or attempts from the slot one team has that illustrate the quality of their chances, nor is shot quality summed up by a team’s shooting percentage. In fact, one of the reasons I decided to sort through some of our data and write this piece was this tweet from Matt Cane (@Cane_Matt) over the weekend: “Doesn’t this boil down to shot quality matters; Sh% isn’t the way to measure it?”

Relying on location data alone is a mistake. We’re finding out from Chris Boyle and, more recently, Steve Valiquette, that screens, puck movement, and specific passes have as much, if not more, to do with shot quality and shooting percentage than simple puck location at the time of the shot.

How the puck gets to area where the shot is attempted is more important than where the shot is attempted from. Shot quality exists and is more complex than location data alone.

As we track more data, I’ve decided to take a deeper look at some of our various metrics and their influence on shooting percentage. Previously, I’ve posted here looking at Sh%G (Overall Shooting Percentage Generated from passes) and a team’s overall Sh%. Today, we’re going to look at updated results for those metrics and add three more to enhance our understanding of how goals are scored: non-passing Sh% (goals on shots that were not preceded by a pass) A1 SH%G (goals scored on shots preceded by a single pass) and A2 Sh%G (goals on shots preceded by multiple passes). These numbers are from all 5v5 situations. All non-passing data was taken from This reflects games played through 1/20/15 and data tracked for the New Jersey Devils, New York Rangers, New York Islanders, Chicago Blackhawks, Florida Panthers, and Washington Capitals.


So, the bolded black line represents a team’s overall shooting percentage and range between 7% – 9%. That’s to be expected. The orange line is a team’s overall shooting percentage from passes, which should look familiar based on my previous articles. Of the six teams we have significant data on, four of them score at a higher rate from passes, and the other two (Islanders and Devils) are 0.7% off their overall shooting percentage.

What’s new here are the next three lines. Gray represents a team’s shooting percentage from a single pass (A1). You’ll notice this is not that successful and lags behind a team’s overall shooting percentage for nearly every team. The yellow line represent’s a team’s shooting percentage from multiple passes (A2). Here is more data on my point above: how the puck gets there. Overall, of the teams we track consistently, they are 3.5% more likely to score on a shot from two passes as opposed to one.

The final line represents a team’s non-passing shooting percentage. To arrive at this number, I subtracted the number of shots and goals we’ve recorded from a team’s total goals and shots in 5v5 situations. What was left were events not generated from passes. I divided the number of goals by the number of shots and arrived at a non-passing shooting percentage. You’ll see that it is in direct opposition to a team’s Sh%G figure, and only slightly better than a team’s A1 Sh%G.

The next chart looks at this data in close situations.


Here, we see a closer relationship between a team’s overall Sh% and its Sh%G, yet the latter is still more successful. The Islanders’ A2 Sh%G is above their non-passing Sh%, as A2 Sh%G becomes the best metric for goal-scoring except for the Devils, who have a slightly better A1 Sh%G.

Back to my earlier remarks: how teams generate offense matters. By breaking shooting percentage down into various scenarios of what precedes a shot (multiple passes, a single pass, or nothing) the data illustrates that “shooting percentage” as we know it, is a bit more complex beast than we may have originally thought. There’s a clear trend showing teams score more when they pass it more effectively. The next question is, why?

To start, we can follow up on David’s work from this past week by looking at the relationship between shooting percentage and how efficient teams are at passing the puck for these same six teams. How we measure efficiency, if you recall, is the percentage of shots generated from passes as a proportion of all shot attempts generated from passes (SAGE). The logic behind it is this: if a team is more efficient at getting shots on goal, they won’t need as many attempts. In the next series of charts, I’ve plotted a team’s overall shooting percentage and their efficiency from a single pass, multiple passes, and overall.

Since our passing metrics are, at the moment, entirely offensive metrics, I also included the total number of shots for each team. The hope is that once our data is available on War on ice, we’ll be able to look at differentials of passing metrics in the same way we look at Corsi, Fenwick, and Shot differentials. On to the charts.


Here, we see an R^2 value of 0.194, which is not great, but evidence of a relationship. However, if we remove the Blackhawks from our group of six teams (an obvious outlier likely based on their poor Sh% from earlier in the season), the R^2 jumps all the way to…


0.985? Wow. Now, since we’re dealing with only six teams already, I’m not getting too excited about this value, but I thought it was worth noting the impact one team can have on our sample size of six. The real value is certainly higher than 0.194, but not 0.985. Let’s move onto Secondary Passing (A2) Efficiency.


So, we see a stronger relationship between the two metrics, indicated by the R^2 of 0.454. This would make sense since I showed above that teams generally score more often from multiple passes. Now, let’s remove the biggest outlier, in this case the Capitals, and see what happens.


The R^2 value jumps to 0.604. So, the value in the original chart of all six teams isn’t too far off. Let’s take a look at overall passing efficiency.


Similar to the raw Sh%G results above, we find overall SAGE nestled between A1 and A2 SAGE in terms of the strength of its relationship to overall Sh%. In this case, an R^2 of 0.332. As we’ve done before, let’s remove the outlier (Hawks again) and see how it looks then.


Another big jump in the R^2, this time to 0.821. So, again, I’m not getting too excited by removing one team from a sample of six, but it’s worth noting the strength of the relationship between the other five when an outlier is removed.

In the next chart, I wanted to look at the shot volume of each team and their relationship to Sh%. If a team simply shot more or less, how would that impact Sh%?


Well, not much. In fact, we’re observing what was discussed in the comments in one of David Johnson’s recent pieces ( a negative relationship between volume and shooting percentage.

So, that might have been a lot to throw at you, but I return to my original point: how teams generate offense and what precedes the shot (no pass, a single pass, multiple passes) matters significantly and has to be a part of any shot quality debate. How teams generate offense and how efficient teams are in their passing matters.

In my next post, I’ll reveal a bit about our location shooting percentages to offer even more detail on how teams are scoring and generating offense. Viva la Shot Quality!

3 thoughts on “Stimson: More on passing data and the shot quality debate

  1. Pingback: The Value of Outliers -

  2. Pingback: GUEST POST: A Call To Action on Crowdsourced Data — How You Can Help Usher In a New Era of Hockey Stats | WAR On Ice: The Blog

  3. Pingback: Taking a pass on new hockey statistics | ALDLAND

Leave a Reply

Your email address will not be published. Required fields are marked *