
According to Merriam-Websterās Dictionary, a network is:
noun netĀ·work Ėnet-ĖwÉrk,Ā a group of people or organizations that are closely connected and that work with each other
There are social networks (Hi, Twitter).
There are professional networks (Hello, LinkedIn).
And, there are team-based networks. In your office and on the court. While these can be seemingly forced, particularly as a function of the lineups that a coach puts together, they are best measured by how a collection of players work together. Lineups are mandatory; networks are created from the successful relationships within a lineup. When that cohesion works in coordinated fashion to achieve a goal, scoring and defending become natural products of the connectivity among that group. If we had to pick one (and only one!) measure to determine how well a lineup collaborated, the choice is clear: itās the pass. More significant than the points scored, a pass signifies teamwork; itās a measure of trust and relationships. Passes get a lineup moving towards a common goal: creating opportunities to score points.
If passing is important (and clearly it is), then the amount of passing a team does should indicate a stronger network cohesion. This has been done before, and this prior work by Andrew Bergmann highlighted how well each starting lineup shared the ball. It was an essential first step of network-based passing analytics.
So letās take that second step, to PASS Networks.

How is this different? For starters, we are examining the entire roster, not just the starting lineup. But more importantly, rather than looking at just total pass activity (or even normalized pass activity), this network is based on Points Generated Per 100 Passes. By normalizing by points scored, we get a sense of how valuable different passes are. Further, instead of displaying all connections across the roster, this network only shows above passes above league average points per 100 (which last season was roughly 54 points).
So as youāre digesting this chart, here are some of the things to pay attention to:
- lines: pass activity weighted by points per 100 passes (thicker means more points) ā notice the direction of the arrows
- nodes: connectivity ā the larger the dot, the more teammates that player delivers passes to
- colors: communities ā players exhibiting common characteristics
Armed with that knowledge, whatās good about last seasonās Warriors PASS Network? First, itās full of large nodes: Golden State had a lot of players making above-average passes for scores. A lot. Second, thereās a fair amount of bidirectionality: two-way communication or, reciprocity. Third, thereās only two communities, which organize roughly into main contributors and reserves. But above all, look at the levels of āsuper-facilitationā from Draymond Green, Andre Iguodala, Shaun Livingston, and (to a lesser extent) David Lee. And, look how many players feed Klay Thompson (and not Steph Curry) at an above-average level!
Now letās compare these historic Warriors to a less-historic team, last seasonās New York Knicks:

The differences are astounding. First, thereās completely unconnected nodes throughout the roster: players who have no part whatsoever with pass activity at above average scoring levels. Second, none of the nodes are particularly large ā they had no āsuper-facilitators.ā Third, thereās three disjointed communities. Sure, some of that is attributable to midseason trades, but plenty of it isnāt.
If we reasonably believe what good and well, not-good, look like, letās fast forward to this season and take a look at Golden State thus far:

Truth, there are currently four communities; this is likely a function of only a fraction of the season completed. Judging by the size of the nodes and the width of the connections (not to mention an undefeated record), the process for getting buckets is strong. Very strong. However, an interesting note: notice how many more teammates are feeding Steph Curry this season, compared to last. This does not mean that Leandro Barbosa is passing to Steph Curry a ton. It does mean that when Barbosa passes to Steph, the result is oh-so well above average.
As for this yearās Knicks, their PASS Network appears to be a step in the right direction:

Still three communitiesā¦but that should be the least important takeaway here. They are feeding each other and working together. Look at all the players assisting to Carmelo, at above average levels!
So what ā how is this helpful? At first glance, it may seem that this is no better than just looking at Assists statistics. I would disagree ā aggregate stats are great, but not at the nuance. PASS Networks show the connectivity (or lack thereof) across teammates. Look back at last seasonās Knicks ā who wouldāve thought Sweet Lou Amundson and Cole Aldrich were such a dynamic duo?! This isnāt just about generally creating shots; networks get you analyzing the best dynamics across the entire roster.
More to come throughout the season.
Data and photo support provided courtesy of NBA.comĀ and data extraordinaire Darryl Blackport.