APBRmetrics

Hey y'all,

For some reason I have read a startling number of articles about, as Crow put it in some thread, the "value of player motion" . First there was these two at nba.com..

http://hangtime.blogs.nba.com/2014/10/2 ... pass-more/
http://hangtime.blogs.nba.com/2014/10/1 ... -movement/

Then there was one at big league insights...

http://bigleagueinsights.com/nba-advanc ... -distance/

And now one at Harvard' sports blog...

http://harvardsportsanalysis.org/2014/1 ... n-the-nba/

Any thoughts on this topic anyone?

I'd like to see the correlation between #passes and feet run with distance from shooter. Might help to also see avg. time elapsed between final pass and shot. Also does distance from man with ball tend to build with each pass or just bounce around? I'd theorize that teams who take advantage of movement tend to pass in ways that builds distance of defender from man with ball and eventual shooter and may shoot faster after receiving final pass but need to see data to prove or disprove.

If you looked at distance from shooter on avg. by team and player by second or portion of shot clock how much difference would you see? Are the differences fairly consistent or is there a zone where the difference in decision making to shoot is most different?

How does ball and player movement track with offensive performance in last quarter, last game, week or month?

Do we know the avg number of passes and player movement for the 5-10 most common play types or shot types? Seems possible and useful to get this background information.

We really need to get player tracking data broken down by "offense" and "defense".

Crow has some good suggestions -- in theory, a motion offense that passes a lot (Spurs) should be getting more open shots.

Now that SportsVu is showing defender distance in the shot logs (see here: http://stats.nba.com/player/#!/201939/t ... shotslogs/), we should be able to take a look at correlation between passes, distance, and "defender dist" when the shot is taken. Hopefully we will have time to play around with that data soon.

I see in your recent article that pace is down. Guess the extra passing and player movement is contributing. Using more clock is a negative on average. Sometimes there is a happy median, not noticed by a correlation.

(That also may be true for 3 pt rate, as indicated in a recent nylon calculus article.)

I looked at some SportVu stats on player motion.

Value of Player Motion: Does Motion Create More Open Shots?
http://bigleagueinsights.com/value-play ... -tracking/

Short answer: no.

Thanks for checking. Breaking distance into offense and defense is indeed still needed.

Another thing that would be interesting to see would be avg distance of all offensive players to nearest defender, % within 4 ft, % greater than 6 ft and avg. distance between players and avg distance from each player and the ball. Maybe spacing is more important than ball and player movement.

Late in my playing career (pickup ball and church league career) I'd intentionally stand still for several seconds, and then make a cut as soon as my defender turned his head. And I'd get a layup or an open three, and my defender's teammates would be cussing him for losing track of the old guy who can't move.

Point is: I think one of the reasons why "motion" is showing up as NOT creating more open shots is because "motion" in the study is distance traveled. But distance traveled isn't really a proxy for motion. What matters is the timing of the motion and its coordination with teammates and ball.

My guess is that the distance the ball travels on offense will probably correlate better with open looks than player movement. Dumping the ball into the post, drawing a double, and then whipping it station-to-station (post-wing-top-wing-corner) probably won't have much player movement, but it often ends with an open look.

BigLeagueInsights wrote:I looked at some SportVu stats on player motion.

Value of Player Motion: Does Motion Create More Open Shots?
http://bigleagueinsights.com/value-play ... -tracking/

Short answer: no.

One issue with this analysis is that you correlate distance per possession with shots over the season. The problem is that teams with longer distance per possession (potentially) also have longer possessions and thus fewer of them, and so fewer shots of any kind over a season. That alone would lead to a negative correlation between the two, which could offset any possible positive correlation brought on by the benefits of motion.

I ran a few versions of analyses that took number of possessions into account in order to get more of a 'unguarded shots per possession' correlation. Motion always had a positive influence, but its statistical significance depends on the exact formulation of the analysis. At the simplest level, there's about a .2 correlation between distance/possession and UFGA/possession, which isn't great but is much higher than what's reported in the blog post.

That's a good point xkonk. Distance per possession vs. UFGA per possession is more of an "apples to apples" comparison.

I will revisit this when we have the new SportVu "shot clock range" data found here:
http://stats.nba.com/player/#!/203076/tracking/shots/

Maybe I'm thinking about this wrong, but aren't those correlations kind of answering the wrong question. It seems to me that we shouldn't care as much whether "do teams that run more get more open shots?", but rather "when teams run more do they get more open shots?" . There are so many potentially confounding variables, like pace (as was already brought up) style of play, player talent, etc, that straight up comparing distance run to open shots (or even % of open shots) isn't the right way to go. Ideally, if we had possesion detail, we could control for team, length of possession, and type of possession (fast break, iso, post up) and see if when teams move more on those possesions they increase their scoring or something.

On a somewhat related note, does anyone know a good place to get time of possession on a game by game level?

How the possession starts has a significant impact on how it ensues, beginning with whether or not a team can run off of it.

Live ball turnover? Assumption is teams always run, but that is largely dependent on how deep the preceding possession got before the turnover. Extreme example - stealing an inbounds pass leads to "running" far more often than deflecting a post entry and gaining possession in the corner with 90+ feet to the rim. Every LTO within that continuum has a varying degree of runnability.

Every other possession start has the same continuum, though less pronounced. Trying to run off AB3s is different than trying to run off C3s, which are different from trying to run off RA misses, which are different than trying to run off RA blocks. Generally speaking, a Restricted Area shot puts the defender behind the ball whether it is made or missed; an RA block increases that gap (provided the blocking team takes possession). Running off a missed FT is much different that running off a missed elbow jumper in the flow of the game.

Unless you are operating on the far ends of the spectrum - a live ball turnover with offensive numbers on one end, a reset halfcourt offense on the other end - player and ball movement is dictated to some degree by how the possession originated.