First Impressions of Player On/Off Distancology

Apr 8, 2015; Dallas, TX, USA; Dallas Mavericks guard Rajon Rondo (9) drives to the basket past Phoenix Suns guard Eric Bledsoe (2) during the first quarter at the American Airlines Center. Mandatory Credit: Jerome Miron-USA TODAY Sports

The following plots will need a bit of time to understand. Trying to simultaneously condense and clarify my previous work on distancology, I might be close to breaking some laws of design. I promise, once you understand what is going on in my plots, they are very intuitive and rewarding. But therefore, I would propose that you read the next few lines very carefully.

1. What are the following plots about?

To win a basketball game, it is generally advantageous to create high percentage shot opportunities for your team, while forcing the opposing team into bad shot opportunities (duh). The NBA.com SportVU data gives us the opportunity to measure a team’s or player’s success in these fields.

The beautiful thing about basketball is that all players can influence the outcome of a possession, without necessarily touching the ball.

In the following plots, I try to describe the influence of each player on their team, by showing information about each player on the floor during at least 400 shot attempts (on offense or defense)[1. Big thanks to Darryl Blackport for the shot log data]. The left half shows the shot frequency of the own/opposing team while a specific player is on the field and the right half shows the shooting percentage.

2. What do the columns and the colors mean?!

Using SportVU data, I separated information from the team shot logs into two-foot increments. So, 1 f00t actually goes from 0.0 feet to 2.0 feet and so on. As there are not many two-point attempts between 22.0 and 24.0 feet, I added those to 21 feet. I also added all shots between 26.0 and 28.0 feet (aka the Damian Lillard range) to 25 feet distance. It is not perfectly separated, but you can mostly assume that 23 foot three-pointers are corner threes and the 25 feet ones are regular threes.

I normalized each column by the NBA average. The problem with showing shot frequency and shooting percentage at the same time, is that they have different variance. Therefore, I multiplied the shooting frequency by 2 on offense and by 4 on defense (as the variance for offense is around 2 times higher than for defense). The leftmost column is the effective FG% for each player – as kind of a summary of shot frequency and shooting percentage.

As an example:

Philly took, in general, about 5 percent more non-corner three-pointers than an average NBA team (dark green 25 feet column on the left half). They also took around 2.5 percent more shots from 2 to 6 feet than an average NBA (bright green 3 feet and 5 feet columns on the left half). This would be considered good, if their shooting percentages from this part of the field were not 5 percent to 10 percent worse than NBA average (bright red and dark red columns on the right half). As an result, Philly’s effective field goal percentage was, for most players, at least 5 percent below NBA average[2. NOTE: All this is not the players PERSONAL percentage, but the sum of everything that happens while the player is on the field]. Only Tony Wroten is a slightly bright spot in this regard with an on-court eFG% of around 3 percent below NBA average.

3. Why the ‘(shot clock <18)’ thing?

I was interested in focusing on half-court situations, so I filtered out all plays that took less than 6 seconds.

4. What does these plots not show

Anything related to turnovers, fouls and rebounding. So, if you would play rim protection like a lumberjack, your defensive plot would look much better than it actually is. On the other hand, players that get to the line a lot (see Harden, James), would have a better true shooting percentage than the effective field goal percentage indicates. But field goal information itself can already teach us a lot.

Examples of what these plots can teach us:

1. Beware of simple on/off comparisons

There are several other ways to show this problem, but when I saw this tweet from John Schuhmann

I directly got reminded of this figure:

So yes, Chris Paul has an amazing on-off differential[3. Having ‘an amazing on-off differential’ could also double as an innuendo for well-endowed stat guys.], but it also helps your differential a lot if you play almost exclusively with either the rest of your starting five or Jamal Crawford (instead of playing with – let’s say – Austin Rivers). The same goes to a slightly lesser extent for Steph Curry, who was on the floor during 88.5% of Andrew Bogut’s and 85.4% of Draymond Green’s minutes. But more on Golden State’s defense during the next paragraph.

2. Great rim protectors should have a visible effect in these plots[3. Unless you have two great rim protectors and the time on the field is divided between them. It’s always tough to be bold when it comes to sports data…]:

The most striking example for this hypothesis is the Stifle Tower:

The opponent’s shooting percentage around the basket while Rudy Gobert is on the floor is more than 5 percent below the league average for shots from 0 to 8 feet. As the French would say: ‘Beaucoup de bras, pas de chocolat’.

While Golden State’s whole first unit is a defensive juggernaut, you can still see Andrew Bogut’s effect on rim protection:

In addition, you can see how the Warriors are really great at forcing mid-range shots (left side of the figure).

On the other hand, DeAndre Jordan and Anthony Davis — two other bigs that are mentioned as DPOY candidates — show no paticular rim effect for mediocre defensive teams:

Side note: It is funny to see Austin Rivers having the best defensive eFG% for the Clippers and the worst one for the Pelicans, a lot of it related to differences in three-point percentages. Which probably tells you a bit about the reliability of plus-minus stats…

3. Rajon Rondo, destroyer of space

This figure was probably the most interesting (or depressing, depending on if you are a Mavs fan or not):

We all knew that the Mavericks offensive rating went down since Rondo’s arrival, but with this figure one can almost physically feel, how three pointers turned into long twos. And if the Mavericks are the Guardians of the Galaxy, then Rondo is most likely Drax, The Destroyer Of Space.