More on Players “Getting Their Own Shot” and Offense Creation

facebooktwitterreddit

Dec 16, 2014; Memphis, TN, USA; Golden State Warriors guard Klay Thompson (11) drives against Memphis Grizzlies forward Tayshaun Prince as forward Draymond Green (23) sets a pick at FedExForum. Mandatory Credit: Nelson Chenault-USA TODAY Sports

Yesterday, in looking at Steph Curry’s ability to shoot threes off the dribble, I lamented the fact that we don’t have the tools to perform a similar exercise for all shots. After all, we still don’t know which shots were “potentially assisted[1. I use the term “potential assist” interchangeably with SportVU-tracked “Assist Opportunities,” a piece of data that’s made a lot of analysis such as True Usage and the Point Guard Personality Test possible.]” aside from the ones which go in.

Upon reflection, I may have been hasty in that assessment. When looking at just threes, I focused on the number of dribbles a player took before the shot, as on a three-pointer, that’s a decent indication of how the shot came about. However, there is another data point tracked for every shot which might be as, if not more, instructive in determining whether the opportunity for the shot was created by the shooter’s own ability, or through set up by a teammate. This other data point is called, in SportVU, “Touch Time” and simply put is the duration the player possessed the ball before shooting.

On made baskets, assists are fairly predictably awarded based on touch time and number of dribbles taken prior to the shot:

This makes perfect sense, intuitively, the longer the eventual scorer has the ball, and the more dribbles he takes the less the pass had to do with the eventual made shot. While number of dribbles clearly matters in the award of an assist, touch time appears to matter more. Further, the effect appears somewhat uniform across shot types, eliminating a main concern with using dribbles as a proxy for shot creation:

One interesting aspect of the above chart is the comparatively low percentage of close shots[3. Shots taken within 5 feet of the basket] resulting in assists at shorter touch times. This is almost surely a result of offensive rebound put backs and perhaps breakaway steals. These sorts of shots have always somewhat clouded analysis of shot creation using percentage of assisted FGM as a proxy. After all, a team’s offense is seldom if ever “throw it at the the rim and get tip-ins[4. The combination of their own iso-heavy offense, the offensive rebounding prowess of Tristan Thompson and Timofey Mozgov along with Atlanta’s struggles on the defensive glass may make this appear to be the Cavs scheme at time in the Eastern Conference Finals.].”

Touch time allows for these “opportunistic” buckets to be excluded from a count of shots a player “creates” for himself in the colloquial use of the term, the “here’s the ball, go get us a bucket” meaning. Of course, merely breaking shots out by touch time results in an estimation of in terms of the kinds of shots players are getting. Some shots are likely being mis-allocated in each direction. Further, the SportVU tracking data isn’t perfect when it comes to tracking possession of the ball[5. In researching yesterday’s article I found a few dozen instances where the system overestimated pre-shot touch time because it was unable to recognize the rapid changes in possession when a player dribbled for a while than ran a quick give-and-go with a teammate. It was only about 30 shots out of nearly 200,000 which were readily identifiable, a very respectable error rate. Still, some stuff is getting missed and mislabeled, and that’s ok for our purposes.].

With all that understood, for purposes of simplicity, 2.5 seconds of touch time before a shot seems to be a natural inflection point.[5. Made shots from all areas dip below 50% assisted at 2.6 seconds of touch time, though three-pointers drop off a little earlier.] By comparing a player’s shots taken on either side of this line, we can begin to get an approximation of the efficiency with which they can in fact create shots.

The applications of this frame of analysis are wide ranging. Before the season, one of the main arguments against Minnesota accepting a Klay Thompson-centric offer for Kevin Love was Thompson’s impending restricted free agency, for a player who was by reputation, reliant on teammates for his shots. And it’s true, he was! Klay was assisted on 94.6% of his threes and 62.2% of his twos last season, by comparison Curry was assisted on 45.6% and 19.7% and league average assist rate was around 84% and 52% of makes from three and two-point range respectively.

Those numbers could represent role as much as anything. With Curry driving the bus, Thompson wasn’t asked to be much more than a finisher. How did he do stepping outside that role? It wasn’t a disaster. On the 25.5% of his shots taken after more than 2.5 seconds on the ball, his eFG% was 41.6%. On his quicker, more teammate-aided shots, he shot 56.6%[8. Interesting to note Thompson’s rough assisted/unassisted split tracked with the overall gap between shot types league wide last season.]. Though not disastrous, the 41.6% was below league average[9. Around 43.5% eFG on “self-created” shots.]. The ball ending up in Thompson’s hands with something needing to happen wasn’t desirable, but it also wasn’t a hopeless possession. Not a franchise cornerstone type offensive player though.

Fast-forward to this year, and Thompson came out of the gate like gangbusters. There seemed to be a categorical improvement in his ability off the bounce. And the numbers agree. Not only is he creating more of his own shots (about 32.1% this season), he’s made them at an above average clip (46.6%). Though he still struggles shooting threess of his own making (I estimated 25% accuracy using yesterday’s “dribbles” method. Today’s touch time version has him shooting 30.4% on self-created threes) he’s taken a big step up from last season when he shot under 15% on that kind of shots, and become FAR more adept at finishing at the rim[10. 58.3% this season versus 47.6% last year on “created” rim attempts.]. Suddenly, he looks every inch of the a legitimate two-way max player and All-Star.

Looking at some of the other “up-and-coming” wing players:

A few takeaways. First, just how much of Andrew Wiggins inefficiency stemmed from being forced to create so many shots for himself before he is really ready to do so? Caveat emptor on Tobias Harris in free agency this summer if a team is expecting a “primary” wing scorer, while Khris Middleton perhaps has some chops in that area. The exclusion of foul drawing[4. There is essentially no public data I’m aware of which would offer much help on that front.] probably makes Jimmy Butler look like a worse offensive creator than he might actually be, while not accounting for turnovers probably makes Reggie Jackson look better. When did Kawhi Leonard become good at creating his own shot? This is new and mildly terrifying.

Perhaps most importantly, the above should make teams lining up big offer sheets for Draymond Green very nervous, He’s a fantastic player, whose versatility on both ends of the court is a huge factor in Golden State’s success. But a team signing him to be a star or cornerstone player might be in for a rude awakening if they expect to put the ball in his hands a ton.

Lastly, a final comparison. “What is going on with Atlanta” has been a big question this entire post-season. While some of it is missing shots, Kyle Korver being defended a little tighter, Al Horford and Paul Millsap being slightly banged up, etc., their offense just hasn’t looked right. It’s been hard to find statistical indicators, but it certainly seems like the play of Jeff Teague has dropped from the the regular season. As Teague is the sparkplug which gets the Hawks’ machinery running, this has created all sorts of problems, but the root cause has been hard to identify. Until now — Teague was a very proficient creator for himself this season, shooting 48.8% on the estimated 71.5% of shots he created for himself. In the postseason, he’s been forced to look for his own even a little more, with 127 of his 163 shots (77.9%) coming after more than 2.5 seconds of continuous possession, while his efficiency has dropped way off, to 41.3% on these attempts. With this loss of effectiveness, perhaps he is also not drawing defenders the extra step needed to give the rest of the Hawks the extra advantage they need, which in turn makes everyone’s looks just a little less open and a little more rushed. If Teague returns to form against Cleveland it could be just the shot of adrenaline Atlanta needs.

Late Edit

The chart below includes shooting splits for all players with at least 250 tracked shots during the 2014/15 regular season: