Inside the numbers: more interesting tidbits from multi-camera tracking stats

Decrease fontDecrease font
Enlarge fontEnlarge font

Andre Miller, Ty Lawson were among top PGs in potential assists by shot metric. (Garrett Ellwood/NBAE/Getty Images)

On Thursday, I wrote about what the Knicks might learn from a fancy multi-camera tracking system installed last season at Madison Square Garden and nine other NBA arenas. The system, called SportVU and run by STATS, LLC, tracks every movement during an NBA game. It can generate an almost infinite amount of data, on everything from how fast a player runs to that player’s shooting percentage from 19 feet away on the left wing after three dribbles to his shooting percentage with a defender less than two feet away.

The subscribing teams — New York, Toronto, Washington, Golden State, Houston, San Antonio, Boston, Milwaukee, Oklahoma City and Minnesota — can look through the raw data themselves and/or have STATS generate specific reports. The folks at STATS gave an exclusive first look at three such reports last week, and two of them supplied much of the data for that Knicks post. A few readers requested more statistical nuggets, so I thought I’d dump a bunch of interesting findings in bullet point style below. Enjoy:

• Something really interesting happened in Denver last season. One report tracked every time a player drove the ball from an area 20 feet or more from the basket into an area 10 feet or fewer from the hoop — an event generally considered a healthy thing for an offense. The data excluded fast-break drives, which is important to note, since Denver ended up leading the league in qualifying drives by a wide margin. The Nuggets averaged 24.5 drives per game, miles above the league average of 14.6 and pretty significantly above the No. 2 mark (Cleveland, 18.5). Ty Lawson averaged 9.1 drives per game on his own, the highest mark in the league, and more than the Lakers averaged as a team (7.2, a league-low).

It is here we must note the sample size issues involved. Denver is not one of the 10 teams that subscribe to the system, so STATS only had data for road games in which Denver faced a subscribing team — 10 in total.

Still: It’s interesting, especially since the data excludes transition chances; Denver played at the league’s fastest pace last season, and when I first saw the data, I assumed those fast breaks fueled Denver’s huge lead here. Some delayed transition stuff probably trickles into these numbers, but they are still worth noting, especially in conjunction with a second set of numbers from a different report.

That second report measured potential assists by tracking every shot, make or miss, that would have led to an assist under NBA standards. It then divided those assists by shot locations, so that we can see which point guards tend to produce which type of open attempts: shots at the rim, long twos, three-pointers, etc. The idea is to isolate point guards who create a lot of the “best” shots, namely layups and threes, and not long two-pointers — in theory, the worst shot in the game.

Now team context is obviously a huge factor here. A whopping 43 percent of Rajon Rondo’s potential assists led to long twos, the highest number in the sample, but that’s not an indictment of Rondo’s creativity. Any team starting two jump-shooting power forwards (Brandon Bass and Kevin Garnett) will rely to a borderline unhealthy degree on long twos.

But Denver’s numbers here are still jarring. Andre Miller and Lawson ranked among the half-dozen “best” point guards in the league by this standard, with 67 percent of Miller’s potential assists and 56 percent of Lawson’s leading either to threes or shots at the rim. Only Jeremy Lin (68 percent) had a higher percentage of potential assists fall in those areas, a fact I suspect Houston’s geeky brain trust is already aware of.

On the flip side, relatively few of Miller’s and Lawson’s potential dimes led to two-point jumpers longer than 15 feet; they had two of the four lowest marks in the league among surveyed point guards.

Lawson is super speedy and Miller is a basketball professor, but there is something broader going on here — something good, considering Denver had the league’s third-most efficient offense last season. The Nuggets attempted the fewest long two-point jumpers per game in the league, a remarkable thing, considering their breakneck pace of play. Their ability to earn heaps of free throws obviously helped, since shooting fouls wipes away shot attempts on the stat sheet, but you don’t earn those free throws by simply chucking long twos.

What’s happening? Lots of things. Denver’s players and coaches are smart, for one. The Nuggets also played fast and small, often using Al Harrington or Danilo Gallinari at power forward, a set-up that fits with a hyperactive drive-and-kick (and repeat) game. Harrington seemed especially determined to limit his shot attempts to open spot-up threes or dribble attacks last season, the latter typically after pump-faking a defender sprinting to the three-point arc.

The Nuggets also lacked much of a post-up threat or a traditional pick-and-roll/pop big man after the Nene/JaVale McGee trade.

Still: Watch the Nuggets next season.

• A point guard who does not fare well by this “potential assists” measure: Brandon Jennings, up for a contract extension before Oct. 31. Only 49 percent of Jennings’ potential dimes led to either layups or threes, a very low number, while 39 percent alone went to long two-point jumpers. There are mitigating factors: Milwaukee didn’t have any dynamite interior finishers among its big man core last season; Mike Dunleavy and Carlos Delfino were the team’s only reliable wing/guard three-point shooters other than Jennings; Drew Gooden, who started 46 games, loves a long two-pointer (especially after a few head fakes).

Still: These numbers dovetail with the notion, visible on film, that Jennings isn’t the most aggressive or creative prober of defenses.

• Another number that backs that up: Jennings averaged just 4.8 drives per game last season (using the 20 feet/10 feet definition). Among point guards who started at least 30 games, only Jrue Holiday (4.6), Darren Collison (3.3), Jose Calderon (2.4) and Isaiah Thomas (4.4) averaged fewer drives per game. Again, lots of things — roster context, team/coach philosophy, individual skills — are at work for each of these guys. Jennings shot just 37 percent on shot attempts that came via these drives, one of the lowest numbers among 91 players with at least 30 recorded qualifying drives.

• James Harden shot 77 percent on shots taken after qualifying drives in 37 recorded games, a shooting mark that’s both insane and the highest of any player who recorded at least 40 drives with any outcome (pass, shot, drawn foul, turnover, etc.). Manu Ginobili, Harden’s double, matched that 77 percent in fewer games, and Steve Nash (78 percent) edged it in a tiny 10-game sample. Keep an eye next season on Kyrie Irving, who (in just eight recorded games) shot 70 percent on drives and recorded the second-highest points per possession mark among all sampled players, trailing only Harden.

• Those who shot poorly on drives included: John Wall (43 percent in 26 recorded games), Danny Granger (48 percent in 14 games), Raymond Felton (46 percent in 13 games), Klay Thompson (39 percent in 31 games), Monta Ellis (47 percent in 37 games) and Derrick Williams (33 percent in 22 games).

• The team-based data on driving jibes well with the eye test, and confirms there are lots of ways to win in the NBA. The Jazz, Lakers and Sixers posted three of the four lowest team drives per game figures based on 12-, 15- and 14-game samples, respectively. The Jazz and Lakers ranked among the league’s 10 best offenses last season with post- and elbow-based attacks that de-emphasized dribble penetration and pick-and-rolls. That worked fine for them, but you can bet both teams understand the value of a little more balance. The Sixers, meanwhile, had an average offense built around low-risk motion sets full of cuts, dribble hand-offs, off-ball screens and mid-range jumpers. They will eventually ask for more from Jrue Holiday and Evan Turner, both of whom are surely itching to do more.

• As mentioned in my New York-focused post, the data on elbow touches confirms what we see in real time: Big men with touch receive the ball in that spot more than any other type of player. Marc Gasol touched the ball at the elbow a shade more than nine times per night in 12 recorded games, the highest mark in the league. The next five: Kevin Love, Pau Gasol, Greg Monroe, LaMarcus Aldridge and David Lee. Each of these players, save Lee, passed the ball on about 75 percent of their elbow touches.

• The players most likely to shoot after touching the ball at the elbow: Kobe Bryant (57 percent of touches), Serge Ibaka (61 percent), Leandro Barbosa (!) 55 percent, DeMar DeRozan (!!) 58 percent, Monta Ellis (47 percent) and Mike Dunleavy Jr. (47 percent). Each of those guys touched the ball at the elbow fewer than three times per game, save for Kobe, who got it there a hair more than five times per game in the recorded sample. Ibaka touched the ball at the elbow just 1.16 times per game in 55 recorded games, meaning it’s clear he’s supposed to shoot when he receives it there, mostly in pick-and-pops. Bryant, DeRozan and Ibaka each shot around 41 percent on these attempts, placing near the bottom of the field-goal percentage rankings.

• Among big men who got the ball at the elbow a lot, the most likely to shoot included: Dirk Nowitzki (40 percent of touches), Amar’e Stoudemire (42 percent), David West (41 percent) and DeMarcus Cousins (45 percent). A reminder: We’re dealing with small sample sizes for players whose teams did not subscribe to the camera system, a group that includes both Nowitzki and Cousins. For those two guys, we’re talking about 20 to 30 shot attempts total, not enough to be considered meaningful. Cousins hit 53 percent of those elbow shots, while Nowitzki hit just 38 percent. Not meaningful!

• The data confirms that there was not a lot of high-level passing on the wing last season in Toronto, one of the subscribing teams. Barbosa recorded an assist on just one percent of his 78 recorded drives last season in Indiana and Toronto, which means he literally recorded a single dime on all of those drives — the lowest percentage in the 91-player drive sample. DeRozan recorded assists on just three percent of his 121 recorded drives, one of the half-dozen lowest figures in the sample. They both ranked toward the very bottom of the assist chart on elbow touches. The Raptors have long been waiting for DeRozan, also extension-eligible now, to emerge as a more savvy creator. The data suggest they are still waiting.

  • Published On 10:03am, Aug 31, 2012
  • Post comment as twitter logo facebook logo
    Sort: Newest | Oldest
    MrRuiThomas 5 pts

    Masterpiece by Zach Lowe.  The data is spot-on.