Here’s a hill I will die on: The primary goal of spacing in basketball is to manipulate a defense into a providing a high quality shot for an offense.
To me, as a former player/coach/scout/analyst/front office specialist, this is intuitive. However, as one Eastern Conference representative responded to me: I’m not sure about that. Spacing is about creating space. Another Eastern Conference representative gave a different response with a mocking tone: I’m sure Steph Curry doesn’t think “better shoot this because they are 6.2 feet apart.” In fact, teams that have pushed back on the notion are still in the belief that spacing can only be achieved by placing all shooters on the three point line and that the defense plays no role on spacing. Other teams have quickly identified that spacing is indeed a defensive-based analysis where offenses are measured based on the defensive reaction to an offensive scheme.
Current Value of a Dunk (and Layup) versus a Three
When we talk about scoring efficiency, we think of effective field goal percentage as the main descriptor. Remember this value provides an unbiased sufficient statistic for points scored from the field. What this quantity helps break down is the value of every type of shot, and serves as a basic premise of “layups and threes” over mid-range jumpers.
If we take this a step further, teams tend to shoot really well when the play results in a dunk. As most teams shoot 85% or better when the field goal attempt results in a dunk; we would expect a play ending in a dunk to have an “expected point value” of 1.7 points per attempt. That’s not too bad, right?
Compare this to a resulting three point attempt. Using the naive method for comparing two’s and three’s, we find that a shooter must hold a 56.7% conversion rate in order to keep pace with a dunker. And 85% is the low number for teams. Here’s the dunking distribution for Atlanta, Boston, Brooklyn, Denver, Detroit, New York, Portland, and Sacramento:
What we see from these teams is that teams effectively only attempt between 3 and 7 dunks per game. The reasons are simple: dunks have a higher propensity to be defended and are opportunistic. The latter part means that dunks are primarily taken when there is little resistance by the defense. In fact, whenever a dunk is on top of a defender it either turns into a highlight of a scorer destroying a defender or a defender picking up a gritty block.
Compare this to teams that hoist between 25 and 35 three point attempts per game, we see that threes appear with less resistance than dunks. If we take an even slightly more in-depth look at the rim attempts; layups are a significant drop off when it comes to conversion rates. As dunks are typically converted 85 – 95 percent of the time, layups are converted only between 54 and 64 percent of the time; with the Milwaukee Bucks and the Toronto Raptors leading the league in conversion.
Here’s six layup profiles of the eight above teams we highlighted for dunks:
We see between 11 and 16 layup attempts per game. At a clip of only 58%, we obtain an “expected point value” of only 1.16 points per attempt. This is indeed significant drop-off from the 1.7 for dunks; but still well above mid-range attempts. When compared to three point attempts, we naively only require a 38.7% three-point percentage to keep pace with layups; which according to NBA Stats a total of 18 teams currently shoot above that percentage.
This suggests that layups+dunks are a slightly better option than three point attempts. This reinforces the idea of layups and threes; but should really be viewed as Dunks and Threes but Layups if you must. Therefore, spacing in an offense should reflect creating lanes for layups and dunks while opening pockets about the three point line for open looks at three. Here’s what I visualize when I speak of spacing:
The above play is a Hammer Action run by the offense initiated by a Pick-and-Roll. The dashed yellow line is the movement of the defender after the weak-side pin is hit. Just think, if the weak-side screen defender decides to help, this leaves a wide open space for a slip for an uncontested layup or dunk. Spacing is created by the screen, not the three point attempt. In fact, if this is executed perfectly, we shouldn’t see a 3PA recorded. If this play is not executed perfectly, we will see the 3PA recorded. And, of course, if this play is well-defended; we will see neither immediate actions.
What this diagram shows is that spacing is indeed manipulating the defense into a high-quality field goal attempt. Spacing is created by screening, passing, and gravity; not solely three point attempts. Let’s take a look at this in-depth at the expense of DeAndre Ayton.
Putting Ayton on Skates
To illustrate spacing, we look at the notorious crossover move by Indiana’s Darren Collison on Phoenix’s DeAndre Ayton (22) during their November 27th match-up. The play was a result of two attempts at a Myles Turner (33) – Darren Collison (2) pick-and-roll action.
During the possession, Collison brings the ball up the court, and the Pacers run a double pull-through with Thaddeus Young (21) and Corey Joseph (6). Joseph goes to the strong side corner as Young goes into the strong side short-corner. The aim is to pull the defense to the strong side of the court, opening up a driving lane along the right side of the key. The goal is to hit a pick-and-roll at the three point line and attack the lane.
In the screen-cap, we see that Phoenix’s Devin Booker (1) ICE’s the screen, forcing the ball into the congested strong-side. Ayton picks up the ball, but knows he can sag back as Collison’s drive will be directly into traffic. Therefore, there’s no driving lane and Ayton is quick to recover on a slip-pass to a rolling Turner.
As there is no space to act, Turner either must force a difficult shot, or bring the ball back out of the paint. Smartly, he does so by kicking back out to Collison, who has now strayed back to the three point line. Turner then chases the pass and attempts the pick-and-roll a second time.
As we see Turner come back up, Booker holds his ground and does not ICE the pick-and-roll the second time. Ayton correctly reads the play and sags to contain Collison’s dribble; except this time Turner pops instead of rolls. This action forces Booker to go over the top of the screen and allowing the ball to come downhill into the lane.
If we compare the two frames immediately after the pick-and-roll; we see that Ayton is in nearly the same spot. He’s properly contained the ball going from right to left. However, his feet in the second pick-and-roll are tight. This means his range of motion will be poor; which is what Collison capitalizes on. Regardless of the feet, Collison’s driving lane is closed. This forces Collison to hit the brakes.
Ayton is unable to control his body due to his footwork, which Collison actually pauses to watch Ayton fall. As Booker recovers in the play, we still see he is not guarding anyone as he’s facing the wrong way to defend a catch-and-shoot to Turner; facing the wrong way and can get nailed with a DHO to Turner; and facing the wrong way to help on Collison.
Similarly, Phoenix’s T.J. Warren (12) is the next man up on the stop. He is facing the play with Thaddeus Young behind the backboard in the (now) weak-side short corner.
Instead of stepping up, Warren takes a moment to process the action. Collison drives. Believe it or not, the sequence between Ayton falling down and Collison leaping up for the layup is 1.92 seconds. That’s a significant amount of time to watch a play unfold.
Here is a situation where spacing is created by overloading the strong side of the ball and using a screen to attack a driving lane. While Phoenix played the play properly (enough), unfortunate luck combined with two bad reactions from Booker and Warren allowed Collison to have no one within 8 feet of him as he went up for a layup. That’s spacing with only two guys outside of the three point line.
Anticipate and React
On way to think of creating (or losing) spacing is the ability to open (or close) passing and driving lanes. This is usually coined as anticipation and reaction. Anticipation is the result of understanding the actions of an opponent and beating them to spots. Reaction is the process of taking appropriate actions after an opponent has initiated a move. Good NBA players are athletic enough to react. Great NBA players are athletic enough to anticipate and react.
In the example above, Ayton was very strong at anticipating the action. However, in the situation with Collison, Ayton’s reaction was a bit slower. One way to measure this is through velocity and acceleration. When I described how Booker was facing the wrong way or Ayton was just “going in the wrong direction,” I was focused on their acceleration and velocity. Let’s see the above possession again, but with the coverage map and the velocity vectors attached.
If we take a quick screen-cap, we see that Booker is actually defending the roll that never comes. And since it never comes, he never corrects himself and heads away from the play; putting him out of position. This spacing is created by Myles Turner.
Mikal Bridges (25) needs to hold his ground as he’s the last line of defense for a kick-out to a wide open Corey Joseph. Plus, being at his height on the play, he has to come downhill to attack Collison; a likely fouling (or futile) attempt on the ball. The reaction time of Warren leaves him outside the opposite end of the restricted area when Collison lays the ball in; over six feet away.
Side Note: Sketching
As a side note, if you’re familiar with sketching, this year has been outrageously rough on the assumptions of the sketching model. If we take a look at the associated velocity plot of the offense, we find that many of the actions will not segment until improper times.
Here, we see that the two pick-and-rolls do not meet the sketching segmentation process’ requirements. Over the previous two years, pick-and-rolls have become increasingly non-traditional in the sense that a screener sets their feet and absorbs contact. In fact, many times, they will decelerate into an oncoming defender, and then glide off without stopping. We actually see that both times in this possession. The velocity plots concur. The closest we come it the first screen being initiated. However, Collison is already gone before Turner comes to a stop and initiates contact. This is due to the BLUE action.
Similarly, on the crossover, Collison’s speed is above the 0.1 foot per second thershold. It’s not until he hesitates to watch Ayton fall that he comes to a stop. What this indicates is that segmentation should also include player interaction. Which, by the way, if you managed to appear at a Baltimore convention I spoke at in 2015 (along with John Urschel, who spoke about Graph Laplacians) you would have seen my segmentation approach; applied to a Spurs-Heat NBA championship game from the year before.
Hopefully you see why I die on this hill and say that spacing is a defensive driven metric. Also, if you’re curious as to how I obtained the data. I spent way too much time between November 27th and December 1st plotting points. The process I use is here: https://www.youtube.com/watch?v=PNYgqWqS1B8&t=7s
I did manage to confirm through NBA stats that the generated summary stats during the segments matched their summary statistics (within .01%). So that’s a plus. But it’s also 9 hours of my life I won’t get back.