Having Some Fun With the Passing Project: Dangerous Duos

Updated: September 9, 2016 at 10:04 am by Loserpoints

Over the past several months, Ryan Stimson’s (@RK_Stimp) passing project has been bringing new information to the public about what types of passing sequences lead to the most dangerous types of shots. Ryan, Sean TierneyGus Katsaros and others have written excellent pieces on how this data can be used to analyze team performance. I’ve written a couple of pieces attempting to evaluate individual player performance based on the data tracked in the project. All of this work is based around the concept of primary shot contributions (PSC), which include all shots and primary shot assists. As I’ve been working with the data, one idea has always followed me. I want to know which two players combine to create the most dangerous shots for their teams. Using the weighted PSC metric that I introduced earlier this week, I can finally take a first look at exploring this idea.

But before we go any further, I need to state up front that this article is more of an exercise in fun than any type of meaningful analysis due to some significant limitations in the data. To start, I don’t know of any way to access TOI for duos. So since I can’t calculate per 60 minute rates, I’m going to present the numbers in two ways. First is per game, which is far from ideal. Using per game numbers results in players who play together more frequently generating higher numbers. Similarly, teams that keep consistent lines rather than blending are more likely to produce more dangerous duos when looking at the data in this way. In an attempt to provide some additional context, I’m also including the wPSC value per shot, which will help capture duos who generated dangerous chances despite not playing together as much. All data in the set is limited to 5v5 and contains only duos who combined to generate at least 0.5 shots per game, which is admittedly an arbitrary number used largely because it’s where the data became reasonable for visualization. In all, that encompasses 562 different duos.

Below is a bar chart that shows all 562 duos and can be filtered by team.

Each duo is measured by the wPSC of the shot attempts they combine to generate per game. As expected by anyone who has read any of the previous work on the passing project, the Sedins are tops in the league, which means that shots where one Sedin makes the primary shot assist and the other Sedin takes the shot have the highest wPSC of any duo in the league. Essentially, they are the most prolific offensive tandem in the NHL based on the passing project data. Next to each bar in the graph is the percentile rank for each duo among the 562 included in the sample, which is provided just to offer context when the view is filtered to show a specific team.

While that chart does a decent job of showing the most dangerous duos in the league, the numbers are skewed by players who spend more ice time together. The chart in available in the second tab, (Duo shots created and wPSC Per Shot) which can also be filtered by team, shows the total number of shots generated by each duo per game on the x-axis and the wPSC per shot on the y-axis. Numbers high on the x-axis indicate that the duo combines to generate lots of shots and numbers high on the y-axis indicate that the duo generates more of the types of passing sequences that tend to lead to goals. Size of the dot indicates wPSC per game (the same measure from the previous table). The dotted lines represent one standard deviation above and below average by each measure. 

While this article is meant to be more of a fun look at trying to assess dangerous offensive duos, the data collected in the passing project has amazing potential to provide insights into tactics and strategy that haven’t been available in any previous public data sets. Some of that work is already being done by the authors mentioned in the first paragraph and some is still under way. So if a quick look at some fun numbers like this has your interested piqued in the project, be sure check out the live stream of the Rochester Institute of Technology Hockey Analytics Conference tomorrow where Ryan and Matt Cane will be presenting new research based on the project. And if you have interest in helping contribute to the passing project, reach out to Ryan on Twitter and he can give you more information on how to help.