The Pass is Prologue — Playmaking Locations

facebooktwitterreddit

Mar 8, 2015; Oakland, CA, USA; Los Angeles Clippers guard Chris Paul (3) looks to pass the ball away from Golden State Warriors guard Shaun Livingston (34) in the second quarter at Oracle Arena. Mandatory Credit: Cary Edmondson-USA TODAY Sports

The most well-developed area of stats-based knowledge about the NBA concerns shooting. We know quite a lot about which shots are most efficient, who the best shooters are and so on. One of the big leaps forward in translating NBA statistics into a more useful format has been the proliferation of shot charts, allowing for a quick look to say not just that LaMarcus Aldridge likes the midrange, but that he prefers the midblock on the left side of the floor:

However, far less is known in a systematic way about how those shots came to be. Sure, we know in broad terms how dependant various players are on their teammates[1. Via a stat like %FGM Assisted — unsurprising the highest perecentages belong to the jumpiest of jump shooters.], as well as which players are frequent providers for others[2. Using traditional Assist% or a more involves stat like the Assist Usage component of my True Usage metric.]. With some legwork, we can even make some broad determinations about the value of different assist types[1. In the past, I’ve looked at the value gained through assisted shot attempts over unassisted ones – much of the Warriors offensive improvement this season can be traced to upping their teamwide playmaking.].

Missing are some key points for analysis. The first is some means of evaluating passing efficiency: most of the data publicly available on “potentially assisted” shots which miss is inferential. For example, from the SportVU passing stats, the Clippers miss 9.1 shots per game directly from Chris Paul’s passes. Next to nothing is known about what kinds of shots those represent. Similarly, where on the floor these key passes originate from is largely opaque. Is there some way to measure if getting to certain spots or areas on the floor is best for playmaking? Do individual players consistently set up teammates for high quality looks in a measurable way, or is it just a combination of teammates skill and shotmaking luck from a given players’ passes?

We still don’t know the full answers to those or similar questions. However, via SportVU tracking data, we can begin to unravel those mysteries. It’s broadly understood that three-point shooting is generally a two-person[2. At least two-person—screening is a whole other kettle of fish in terms of tracking, valuing and evaluating.] job. For example, catch-and-shoot threes (generally assisted shots) are have a far higher average level efficiency than three-point pull-ups (generally unassisted.[4. 36.4%-30.7% through March 7 in favor of catch-and-shoot, per NBA.com] But even that is a little messy, as better shooters have learned to instead of pump-faking a closing defender and stepping in for a long two, to take one dribble to the side while staying behind the arc. This is technically a “pull-up” shot, but is also still an assisted shot. In any event, the difference between assisted and unassisted three point shooting is pretty stark, but the efficiency gap exists all over the court:

L: Percentage on assisted/potentially assisted shots taken from each zone alongside percentage of all shots taken from each zone potentially assisted. R: Shooting percentage on unassisted shots from each zone. Visualization by Matt d’Anna

From every area of the court, assisted shots are simply better shots. Of course that alone isn’t particularly helpful in terms of understanding how a team can achieve the goal of getting these better looks? A first indication is maybe to look at where these key, “potential assist” passes are coming from on the floor:

Percentage of assists thrown from each area on the court alongside FG% on potential assists from each zone. Visualization by Matt d’Anna.

It would be impossible to give chapter and verse on everything which could be teased out of this data and keep this post to readable length, so I’ll just give some highlights.

1. Three point shooting is incredibly cooperative: Just under 80% of all three-point attempts came from potential assists, while over 83% of makes have been assisted. Assisted three-point shots are being made at a 36.7% clip while unassisted shots from deep are only going in 28.1% of the time. Even though, on the whole, threes are pretty efficient shots[5. NBA average three-point shooting is the equivalent of about 52% shooting on two-pointers, all else being equal.], that efficiency has been earned, as simply jacking off-the-dribble threes probably isn’t a recipe for much success. Unless of coure you have Steph Curry who is shooting over 40% from three even though barely over half of his makes are assisted. When commentators talk about Curry breaking defensive schemes, that’s it in a nutshell — a terrible shot for virtually every other player in the league is a devastatingly effective shot in Curry’s hands.[1. The same largely holds true for James Harden, who is shooting 38% from three despite only 49% of his made threes coming via assist.]

2. The above trend is accentuated in the corners: 92% of attempts and 94% of makes from the corners are assisted. This aligns with earlier findings that corner three-pointers are more accurate more because of openness than closer distance to the hoop. The corner is a bad spot from which to run an offense (as partially indicated by the comically low numbers of key passes originating from the corners in the second graphic above), so the ball tends to find it’s way there when a shooter is open and not so much at other times. This also sheds some light on why star offensive players tend to not shoot so many corner threes.

3. The paint outside the restricted area is a strange place: Fertile ground for assists. Dead zone from individual scoring. It’s also the area where the lowest percentage of shot attempts are assisted. In a way this represents the modern notion that post-ups are much more about facilitating than efficient scoring, as most “post-ups to score” end with non-RA type shots[5. Hence the relatively low efficiency of post up possessions.], but the threat of a single defender getting overwhelmed into giving up a layup might draw a double and allow for a kickout to open shooters or further ball movement.

4. There’s plenty we still don’t know: It is overwhelmingly likely that both the high rate of assists generated and the efficiency of those shots on passes from the above-the-break three point area represent those areas being used as sort of a midpoint in an offense, a link between the initial defense-drawing movement and an open shot in another area of the floor. That or defenses are massively overreacting to the “threat” of a pull-up three pointer. But my money is on the former, and the further away from the eventual shot an offensive action gets, the less data currently available.

In addition, there’s some interesting nuggets to be pulled from individual assist and passing data, but that’s best left for post later in the week as this has already rambled long enough.