Readers of this post will probably be aware of the passing project headed by Ryan Stimson (@RK_Stimp). The idea behind the project is to track the passes that lead to a shot in NHL games. Volunteers track the players making the passes, the player taking the shot, the zone locations of the passes, the shot types and pass types for each shot in a given game. With this granularity of data, we can then begin to piece together which types of passing sequences are more likely to lead to goals than others. Ryan has already done some amazing work in this area over at Hockey Graphs. Reading those articles before proceeding with this piece will give a better idea of the context in which this piece is written. His first piece on playing behind the net can be found here and his second piece on neutral zone play can be found here.
Ryan’s article on playing behind the net is the foundation for this piece. In that article, he looked at what can be learned from the passing project data about strategy in the offensive zone. To do that, he looked at two types of passing sequences. He examined shots that follow passes from behind the net and shots that follow low to high passes. The most important part of his findings in that article was that those passing sequences are repeatable at a team level and thus can be considered a tactic. Basically, it appears that certain coaching staffs encourage playing from behind the net or from low to high more than others, which makes intuitive sense. And when intuitive thought and data driven analysis align, confidence in the conclusion is high.
The passing project tracks several specific types of passes that can occur in the offensive zone. Behind the net passes are one type. Volunteers also track passes back to the point, royal road passes and passes occurring directly from a faceoff. All other passes fall into a more general category and are tracked based on zone and location. Ryan’s piece on offensive zone strategy clearly defines behind the net passes. Passes back to the point is a more subjective term but is used when a player passes the puck from deeper in the offensive zone back to the blue line. Most often, these passes go to defenders. Passes directly from faceoffs occur when the player taking the faceoff gets the puck to a player who takes a shot. Royal road passes occur when the pass crosses an imaginary line that extends straight back from the center of the goal to the point even with the top of the faceoff circles. That line is illustrated below. The creator of the royal road concept is former NHL goalie Steve Valiquette. Shots that follow royal road passes are more dangerous because they force the goaltender to move laterally which will generally create openings for shooters.
As a first step in examining the shots that follow each type of pass, the chart below shows a random sample of shots following each pass type in the offensive zone. Each dot is a shot, the blue dots are non-scoring chances and the orange dots are scoring chances.
The point of this chart is fairly obvious. Shots following royal road passes and behind the net passes are more likely to lead to scoring chances than other types of passes. But to put actual numbers to this visual representation, the following chart shows 5v5 shooting percentages for each shot type if the shot makes it on goal.
Again, the numbers tell a clear story. Passes from behind the net and royal road passes lead to more dangerous shots with the shooting percentage following royal road passes at almost 28%.
Ryan’s work on passes from behind the net clearly established that those passing sequences are not only repeatable at a team level but also significant predictors of goals scored. The first step in this follow up analysis is to repeat his process and apply it to royal road passes. To coincide with his analysis, the sample of teams is limited only to those that have at least 30 games tracked in the project. To measure the repeatability of royal road passes at a team level, we split each team’s games in half and compare the number of royal road passes in each set of games. The result of this initial analysis is that at team level, royal road passes are not shown to be repeatable. The r^2 value is only 0.03.
One conclusion that could be drawn from this is that royal road passing is essentially random and therefore not worth studying further. However, other explanations for the lack of repeatability in this sample are viable. One is that the sample size is not large enough. Royal road passes are low frequency events. They occur even less frequently than behind the net passes. As the sample size of the project increases, the repeatability at the team level could improve.
Qualitative factors could also be impacting the data. Teams tend to collapse to the middle of the ice defensively and that makes it difficult to complete royal road passes. Defensive strategy could be playing a part in decreasing the quantity of royal road passes and thus, making it difficult to collect a sample large enough to show repeatability. Another explanation could be that teams don’t encourage royal road passing because cross ice passes increase the likelihood of a turnover and a potential breakout going back the other way. Whatever the explanation, the data does not support the idea of team level repeatability for this type of pass. However, a lack of repeatability at the team level is not a reason to give up on examining a type of pass that leads to a 28% shooting percentage.
If royal road passing is not repeatable at a team level in the sample, the next logical step is to go one step further and look for repeatability at a player level. In order to do this, we follow the same split half analysis approach. But instead of limiting the sample to players on teams with 30 or more games tracked, the sample is limited to players who have at least 60 minutes played in each half of the split half analysis. TOI was pulled from corsica.hockey.
Below are the charts with the correlations for team level royal road passing, player level behind the net passing and player level royal road passing. Also included is a new measure called Dangerous Primary Shot Contribution, which consists of Primary Shot Contributions where the primary assist is either a behind the net pass or a royal road pass. The name is derived from Ryan’s earlier work, which identifies shots and primary shot assists as Primary Shot Contributions (PSCs). And based on the shooting percentages at the beginning of this article, royal road and behind the net PSCs are the most dangerous PSCs. This analysis only includes the passing portion of DPSCs so the measure is referred to as DPSC A1. Each of these statistics at the player level is calculated as a per sixty minute rate stat.
What we see in these scatter plots is that while royal road passing is not repeatable at a team level, it is repeatable at a player level. And that suggests that at least part of the issue with not being able to observe repeatability at the team level is the sample size. Behind the net passing also shows as repeatable at a player level. And finally, the new measure, DPSC A1 is also repeatable. All three player level values have a p-value of less than .0001 meaning that these results would be unlikely to happen by chance.
The rest of the article will look at the teams and players who most frequently make these types of passes as well as which players frequently take the shots following these types of passes. Here is the first chart, which shows team and player breakdowns of just the shot assists portion of DPSC. The players who excel by this measure could be considered high end playmakers because they frequently create the most dangerous types of shots. The numbers in this chart are per game because compiling per sixty minute rate stats with TOI restrictions at a team level in this format would distort the team totals. A table with per sixty minute rate stats for each player is available at the end of the article.
Hopefully, the chart speaks for itself (otherwise it has no point) so I won’t spend too much dictating what it represents. But interestingly, the team with the most royal road and behind the net passes in the data set is the Vancouver Canucks. That seems surprising given that they weren’t particularly good last year and definitely did not seem like an offensive powerhouse. However, they do still have the Sedins. And those two alone are enough to drive the Canucks to the top of this chart. Not surprisingly, Henrik (aka “Passy”) Sedin leads the Canucks and the entire NHL in dangerous primary shot assists. And before we move on to the shots portion of DPSC, I’ll give you one guess which player is receiving all of those passes from Henrik Sedin.
If you guessed (aka “Shooty”) Sedin, congrats to you. He has one of the highest DPSC shot rates of any player in the league. And finally, here are the DPSC totals on a per game basis.
For each team, players widely considered to be the best offensive player on the team frequently lead the team in DPSC, which is a good sign that DPSC is a reasonable measure of offensive play. One of the goals of the passing project is to identify creative offensive players and differentiate that particular skill set from other more grinding styles of offensive play.
The passing project allows for an incredibly detailed level of analysis. Examining behind the net passes and royal road passes is just one facet of what the data can offer. As the data set grows, we will continue to learn more about the types of passing sequences that lead to goals and which players and teams excel at those types of sequences. As a closing for this article, the below chart allows you to see the DPSC numbers for every player in the data set. The numbers in the table are calculated as per sixty minute rate stats, which is the best way to compare individual players. The table allows you to sort by each field and filter by position or team.