Offensive Crashing

Back in high school, it behooved our team to “keep one man back” on offense. The thought process was simple, if the defense were able to get out into transition, our team would at least impede their progress towards the basket with the hopes of them settling into the half-court offense. Sometimes it worked. In college, a similar approach was used, but instead our teams ran four-guard offenses (small DI and DIII) and we employed the “no one crash on rebounds” approach in an effort to stymie all fast break attempts. Sometimes it worked.

The challenge of offensive rebounding is that there are two antipodal thought processes at war with each other. Do we attempt to extend the possession and give ourselves a second chance, or do we forfeit the second chance opportunity and focus on limiting fast break opportunities?

Basic Proxy: Fast Break Points versus Second Chance Points

As a simple proxy, we can look at fast break points against second chance points. While this is not a one-to-one relationship thanks to turnovers, it gives a little insight as to how a team crashes.

Screen Shot 2019-11-18 at 9.26.40 AM

Distribution of Second Chance Points against Opponent Fast Break Points.

 

Here we see that teams would potentially tend to play a more conservative style. We see that teams such as the New York Knicks lead the league in second chance points per game, while they are 21st in opponent fast break points. This is in part due to two primary factors: the Knicks are currently the one of the worst shooting teams (.482 eFG%, 29th / .422 FG%, 30th) and yet have the best offensive rebound percentage (.263, 1st). This equates to not only more chances at obtaining second chance points, but being effective enough at securing those opportunities.

To put this in perspective, the Knicks have had 167 offensive rebounds against 649 missed field goal attempts.

In contrast, the Toronto Raptors currently post the worst offensive rebounding percentage in the league at .171, through a total of 92 offensive rebounds. While the team does shoot significantly better than the Knicks (.455, 15th), the Raptors still don’t pressure the glass as much. In relation, the Raptors have one of the best opponent points off fast breaks numbers in the league at 11.3 (3rd).

Screen Shot 2019-11-18 at 9.29.34 AM

Net-Zero boundary is included.

 

Including the net-zero boundary, we see that the New York Knicks and the Sacramento Kings are hedging towards the “Fantastic” region; despite neither of these teams being all that “Fantastic” so far this season. It’s relatively easy to conclude the simple proxy is not indicative of on-court performance.

For instance, Toronto is an 8-4 team despite having the 4th worst differential (-2.8 points per game) between opponent fast break points and second chance points. In fact, there are a total of 18 teams above the “break-even” line. Again, this is due to some mixing of fast breaks initiated by turnovers.

That said, of the 12 teams in the “net-positive” region, six teams have losing records and six teams have winning records.

But what we start to takeaway from this simple plot is this: good teams don’t necessarily get second chance points. However, we do see a trend that effectively states that as second chances points increase, opponent fastbreak points increase.

Running a terrible linear regression on this data, we find that there is a shoddy R^2 of 0.22 (thanks in part to New Orleans, Memphis, New York, and Sacramento) but also  significant positive trend. This is a decent initial first step in investigating the impact of crashing.

Use Case: Crash and Get Beat

As a simple use case, we take a look at a particular play in the Atlanta Hawks versus Los Angeles Lakers game from last night (17 November 2019). In this play, Trae Young uses a screen from Alex Len in an effort to shake Kentavious Caldwell-Pope as he brings up the ball. Caldwell-Pope jumps the screen, allowing Len to rescreen Pope as Young throws a right-left crossover and pulls up to shoot over a moderate contest from JaVale McGee.

As Young releases the ball, three of Young’s teammates are positioned outside the arc, while Bruno Fernando is slightly inside the arc. As Young takes this 25 foot three point attempt, no Atlanta Hawk player is within twenty feet of the basket. The probability of obtaining such a rebound should be considerably low.

Screen Shot 2019-11-18 at 10.13.36 AM.png

Len begins to crash. Fernando is being tagged by Davis in case he crashes. 

Despite this, Len crashes from 28 feet out. Fernando trails and makes it a mere five feet as his man, Anthony Davis, absorbs Len in the paint. Given the position of Young’s three point attempt, the ball is expected to travel between 2 and 8 feet, slightly to the left of the basket. Here, Davis has attempted to nudge Len out of the play and leave the entire paint open to LeBron James and Danny Green. Davis misses the box on Len, as Len hunts out the left-of-basket rebound.

Screen Shot 2019-11-18 at 10.13.58 AM.png

Davis absorbs Len, ball is indeed falling left. How many players are in frame…

As the ball falls short, both Evan Turner and Allen Crabbe start to trot back on defense. As they fixate on the ball, they’ve failed to notice that Caldwell-Pope has leaked out on the fast break. As James grabs the rebound, Turner has barely cleared the break in transition defense and Crabbe is aligned with McGee, who is still roughly 72 feet from the Lakers’ basket.

Screen Shot 2019-11-18 at 10.18.12 AM.png

Caldwell-Pope is gone. All that’s left is for LeBron to channel his inner CC-Sabathia and throw a full court strike.

It’s not on camera (it is on Second Spectrum) but James notes that Turner is at the three point line, Fernando is still seven feet from half court, as Caldwell-Pope is ten feet beyond half court. James hits Caldwell-Pope with a full-court pass for an easy dunk.

Screen Shot 2019-11-18 at 10.22.12 AM.png

There are no offsides in basketball.

 

What this use case shows is the coach’s rebounding nightmare. This is not a commonality, but it encapsulates the conscious decision to rebound or retreat. Most times, it’s a simple decision: shots close range will already have rebounders jostling for position. It’s difficult to leak and go unchecked. With 45.1% percent of field goals attempted coming within ten feet, we see that effectively half of possessions already have non-crashing rebounders in place.

Relationship Between Shot Distance and Rebound Distance

To hammer the use cases’ final point home, we used 2018-19 NBA data to identify the average distance of a rebound given the distance of a field goal attempt. By plotting the results, we obtain a rather wonky-looking plot.

boards

Average rebound distance given the distance of a field goal attempt.

 

Here, we see that dunks tend to become rebounds roughy 5-6 feet from the basket. This occurs primarily due to the fact they are dunks, and a defender is typically contesting the dunk. At this point, a secondary player obtains the rebound.

As the distance stretches out to 2-3 feet, we obtain contested jumpers, layups, and hook shots. These tend to fall again to secondary defenders, who are moving within the flow of attacking the basket. The force of a miss shot from this distance is also much less than a missed dunk.

After this, the rebounding distance trends downward as field goal attempts push further out into the midrange. This is expected as time of flight increases for these attempts and there are typically rebounders in position in the lane. Also, we will comment on a midrange phenomenon that occurs later on.

Once we get out to three point range, average rebound distance increases.  It’s not noted in the plot, but variance of rebound distance also increases. Where it is effectively 1 foot for close attempts, the value is closer to 2 feet for three point attempts. The rebound distance effect is a dual between force of the ball hitting the rim and spacing that is typically confounded with three point attempts. That is, less rebounders in position.

Using this plot, we begin to understand that crashing is more of a question surrounding three point shooting than it is for midrange attempts.

Distribution of Rebounds Give Shot Locations

To better understand the “probability of a rebound” we need to start asking questions such as “where do rebounds go?” In this case, we can build what is called a rebounding plot. Here, we segment a set of field goal attempts from the “same” location and mark where the rebounds fall.

Let’s take, for instance, attempts from a particular 1-foot-by-1-foot box on a three point attempt near the top of the key:

boards1.png

While we see a few rebounds trickle out to the midrange and three point line, the bulk of the rebounds fall within the restricted area, with a bulge towards the left (your right) of the hoop. The sharp flat jut out to the right of the hoop (your left) are short attempts. This is a mirror image of the Trae Young use case, where the first dot outside the paint along the baseline is the position where LeBron James secured the rebound.

Now if we move that shooting position slightly in:

boards2.png

 

We find the majority of attempts fall short. This is because majority of these attempts are pullup jump shots, which tend to be short when missed. It almost makes no sense to have a weakside crasher in this situation as over ninety percent of rebounds are falling short on the strongside. Crashing this attempt leads to a weakside leak and a potential fast break. It is here we see that ~20 feet jump shot leads to a short rebound.

As we come into the midrange, we find that shots are peppered about the key:

boards3

And as we move even closer in:

boards4.jpeg

Here, we are fighting with floaters, which tend to miss opposite side of the hoop as they tend to it the back of the iron.

Given these examples, we can start building a “probability” of where the rebound will go. Conditioned on player position when the field goal is attempted, this gives us a prior probability of how likely a team will rebound the basketball. From there, we can look at the effect of crashing by computing the posterior probability of getting a rebound observing where players end up when the rebound is secured and tallying who secured the rebound.

We can also perform a similar analysis on starting fast breaks. By computing the probability of a fastbreak igniting when a team crashes versus not crashing.

The resulting odds ratio then allows us to identify the capability of a team crashing. Running this across the entire league on three point attempts for the 2018-19 NBA season, crashing is actually a negative return on investment. Meaning that crashing tends to indeed ingite fast breaks. And if a team is poor in transition defense, this can spell disaster.

No-Man’s Land / Zone of Death

Across the coaching staffs that I have interacted with over the years, there is one piece of transition defense philosophy that is shared despite having different names. And that is: get back on defense.

In particular, no leading transition defensive player should be in the back court when the ball crosses their own three point line. This area has been referred to as No Man’s Land and Zone of Death. Basically phrases certain assistant coaches have used to express being a poor location to be on defense. I have even heard it called the “Horseshit Area.”

blazersNoMans

“No Man’s Land” (green). Admit it, my subpar Paint skills are amazing.

In the Hawks use case from above, this is exactly the problem encountered: Young is the shooter and falling forward on the attempt. Len actively chooses to crash, leaving Caldwell-Pope wide open to leak. Turner and Crabbe fail to get out of No Man’s Land in time.

As James’ pass crosses the three point line, Turner is stuck deep in No Man’s Land and Crabbe (who is not in frame) is still three feet inside the region. By the time the ball crosses half court, Crabbe has managed to take two steps to get to three feet on the correct side of the court, but he’s already too late.

Screen Shot 2019-11-18 at 11.09.48 AM.png

Where this transition defense ultimately breaks down is due to Allen Crabbe’s indecision to crash or retreat. It is not only his man who secures the rebound, but he is caught in transition defense’s No Man’s Land. If Crabbe crashes, he makes the pass from James much more difficult to attempt. It may actually have forced James to put the ball on the ground.  If Crabbe doesn’t crash, he has to recognize that he’s gotta do everything to get across half-court, as opposed to tracking the field goal attempt until it’s too late.

Coaching Drill: Combating the Zone of Death

One drill that I’ve seen across multiple pro teams is the “transition defense” drill. Of course, that’s a vague name. So Let’s break this down. In a 3-on-3 setting, defenders are stationed in semi-typical positions on offense. Offensive players are positioned along the baseline; coach holding the ball.

Screen Shot 2019-11-18 at 11.24.14 AM.png

When the coach throws the ball to a random offensive player, the corresponding defensive player must run and touch the baseline as the other two defensive players retreat.

Screen Shot 2019-11-18 at 11.26.37 AM.png

There are three objectives to the defensive retreat. First, the teammates must stop transition. Second, the teammates need to communicate properly. Third, the two primary defenders must clear No Man’s Land:

Screen Shot 2019-11-18 at 11.29.14 AM

The two defenders must retreat beyond half court (green line) before the offense crosses the hash marks (purple line).

If the defense is unable to do that, the offense is awarded a point on top of what they can score.

League-Wide Trend

Over the last 15 years, we have started to see the decline of offensive rebounding percentages.

boards5

While last season saw a little bit of an uptick (.229 compared to .223), it’s predominantly crashing from the three point line that has begun to decreases dramatically, bottoming out at .190 last season. This is almost in lock-step with the league trend of retreating as the remedial math tends to back this up:

If a team makes 35% of their 3PA and crashing produces a 24% chance on securing a rebound with a following .55% chance of scoring 2 points; but yields a 30% chance of giving up a transition with 80% chance of scoring 2 points, the expected gain of crashing is -.1404 points. This seems arbitrarily small, but suggests for 30 3PA’s in a game, we will tend to give up an average of 4.2 points if we crash consistently. To put this number into context, 51 of the first 188 games (27%) of the 2020 NBA season have ended regulation within 4.2 points.

That said, the anecdote is not to retreat all possessions. But rather better understand how to tactically crash. That is, design schemes that not only create space for shooting, but also place players in position to rebound. Similarly, designing schemes for retreating when appropriate such as tagging transitioning offensive players and chasing if the tag is too far.

In a game of slim margins, a team cannot afford to let opponents get offsides in transition consistently. Here’s one symptom of such a case. It’s up to the teams to plan accordingly.

2 thoughts on “Offensive Crashing

  1. Pingback: Offensive Crashing — Squared Statistics: Understanding Basketball Analytics by Justin Jacobs

  2. Pingback: Extending Possessions: Geometric Distribution | Squared Statistics: Understanding Basketball Analytics

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.