One of the simplest offensive plays in basketball is the pick and roll. The philosophy is relatively straightforward, a ball handler waits to receive a screen from a teammate and reacts accordingly. If the teammate establishes position, gets low and wide, and makes proper contact with the ball handler’s defender, then the defense is forced to make a choice from a plethora of options. To gain an edge from this play, the offense is looking to obtain separation from the ball handler’s defender and pushing the defense into a quick decision it may not be prepared to handle.
Case Study: Oklahoma City Offense vs. Golden State Warriors (2016 NBA Playoffs)
During the 2016 NBA Playoffs, the Oklahoma City Thunder executed pick-and-roll scenarios to near perfection in their first four games against the Golden State Warriors in the Western Conference Finals. For example, consider this Kevin Durant – Russell Westbrook action on the right wing.
Durant sets a screen on Stephen Curry of the Warriors, allowing Westbrook the chance to go over the top of the screen and into the lane. Understanding that Westbrook is a premier rim attacking guard, Harrison Barnes (#40) of Golden State hedges out to deter Westbrook from attacking the lane. The result is leads to Durant popping to the wing, uncontested. Westbrook passes to Durant; leading to a basket.
In a second play Steven Adams (#12) sets a screen for Kevin Durant’s man, Harrison Barnes, at the top of the key. This time Adams’ defender, Andrew Bogut (#12), sags back at the free throw line.
The difference here is that Bogut understands that Adams is not a threat to shoot from outside and waits to cover Durant. The problem here is that Bogut never engages Durant and continues to sag, goading Durant to shoot an uncontested mid-range jumper. Durant obliges for two more points.
In a third example, we take a look at Nick Collison (#4) screen on Cameron Payne’s (#22) defender, Ian Clark of Golden State (#21). Understanding that Collison is not a premier perimeter shooter (like Adams), Marreese Speights (#5) sags on defense, preferring to engage Payne in the paint.
The result is similar to Bogut’s strategy, as Payne shoots from the free throw line; this time somewhat contested compared to Durant’s earlier attempt. Despite this the result is the same.
So the question is this: what did both teams do right and what did both teams do wrong? The answer isn’t a simple black-and-white response. However, we can construct a methodology to measure how effective the Warriors’ defense was against the Thunder during these four games.
Sideline Screen and Roll
The above three examples are a common pick-and-roll play in the NBA. Here, the screen comes from the center of the court, to create space for the ball handler who is currently on the wing. Most of the time, the court is cleared with the other three offensive players on the weak side, thus turning this play into a typical pick-and-roll in a two-on-two format.
Taking advantage of the rules in the NBA, a defender cannot be posted in the lane without actively guarding a player for three or more seconds. The defender may be on the opposite side of the floor from the man he is defending, however, this means the defender will more than likely be 20 or more feet away from his man; which is usually not a good idea. Hence, the sideline screen and roll commonly starts in this position:
In this diagram, we see that the strong side of the court is in two-man action, with the three other offensive players stretching the court. If any of the defenders move into the lane, they are only allowed to sit in the lane for “2.9 seconds” before being forced to either exit the lane or actively guard an offensive player foolish enough to enter the lane. Failure to do so results in a technical foul and a free throw for the offense.
There are a plethora of decisions for the defense here. Let’s tackle the four primary methods for guarding the pick-and-roll.
1. Double Team
If both defenders jump the ball handler, this leaves the screener to be free. If the screener is a “Flex-4” style of player, such as Kevin Durant or Dirk Nowitzki, or a “Spot Up Big” like Kevin Love, Serge Ibaka, or Blake Griffin (who took roughly 2 3PA per game last season!), then the defense is in trouble and likely to give up baskets. This was indeed the case in example one above.
2. Sag on Defense
If the screener’s defender sags on defense, then the ball handler’s defender is forced to fight through the screen and recover on defense. This defensive posture coaxes a ball handler to shoot a longer distance shot while protecting the lane and being in better position for a rebound. The hope is that the ball handling defender is quick enough and recover on defense and deter a field goal attempt.
The difficulties with this defense, despite being quite popular, is three fold. First, if the ball handler is a premier shooter off the dribble, then they get a wide open look. Second, is the ball handler is a premier rim attacker, this puts the defender in a precarious position as they are left one-on-one with the ball handler. Third, the screener becomes an open roll man and the action turns into 2-on-1.
In the examples above, this happened twice, both resulting in the first option (open look mid-range).
3. Hedge on Defense
Instead of sagging, the screener’s defender may opt to hedge the ball handler. This motion pushes the ball handler off their attack off a screen and away from the basket. This action buys the ball handling defender an extra fraction of a second to recover on the ball handler.
This defense is extremely popular in collegiate basketball but leaves for a wide open roll from the screener towards the basket. Typically this will result in a look from the post or mid-range; rarely an attempt from beyond the arc.
This method is one of the least desirable methods of defense as it commonly invokes mis-matches on the court. Typically teams will force either a small defender onto a bigger offensive player, a slow defender on a quick ball handler, or a high value player into a high-propensity fouling situation.
The BLUE defense (also known as ICE) is a different form of defense that utilizes action on the weak side; allowing for the defense to be in better position. The action is simple, there are two primary positions on the weak-side of the court. These are the “Two-Nine” and the “Nail.”
The Two-Nine position is the lane. It is called the “Two-Nine” to represent the 2.9 seconds a player is able to squat in the lane without actively defending a player. Anything after 2.9 seconds, a defensive three seconds violation is called; resulting in a technical foul and a free throw against the defense.
If the defender is actively guarding an offensive player in this region, the defender may be in the lane for as long as they please. Otherwise, they must clear the lane to reset the defensive three second count. More importantly, if a smart rotation occurs, a defender may slip across the lane and be a third defender to cover the pick-and-roll for a brief period of time.
The nail position if the location just above the free throw line. It is given this name because there is an actual nail located at the center of the free throw line at each end of the court. If you found yourself playing basketball prior to the 2000’s, it would have been typical of your coach to instruct you to align your body with this marking on the court to shoot free throws. And if you were anything like me, a right handed shooter with a herky-jerky release, you probably stood a little to the left of this thing.
This nail location, however, serves as defining the strong and weak sides of the court on defense and is not subjected to defensive three seconds. This a primary help-side location.
BLUE / ICE Action
The action in BLUE is to force the ball handler towards the corner and hedge with the screener’s defender. This action forces the basketball away from the middle of the court, limiting damage that can be done by a ball handler.
This action still has flaws, but allows for easier recovery from the on-ball defender. The only actions the ball handler can take in this situation are to hoist a contested shot, pass to the screener away from the basket, or dribble out and reset the offense. None of these situations are guaranteed missed field goal attempts or wasted possessions, however, it forces the offense to spend and extra few seconds and helps lower effective field goal percentage as shooters are contested more frequently. In fact, SB Nation gave a how-to guide on this in decent detail back in 2013; with the exception of smart rotation action.
All of BLUE: Smart Rotations
However, recall that we led this section with the terms of Two-Nine and Nail. Here’s where BLUE becomes dominant. Let’s consider Tom Thibodeau’s “weak side” defense on BLUE.
In this sequence, the Houston Rockets are attempting a sideline screen with Omir Asik (#3) to free Tony Douglas (#15). Chicago Bulls’ Nate Robinson (#2) moves weak side, forcing the screen to push Douglas to the baseline. Joakim Noah (#13) drops, acknowledging that Asik is not a premier perimeter shooter.
However, Asik is capable of rolling to the elbow, taking a pass that splits Robinson and Noah. The question is why not? As Patrick Patterson (#54) of the Rockets moves to the weak-side short corner, Taj Gibson (#22) holds the Two-Nine instead of chasing Patterson. This motion is the beginning of a smart rotation.
A smart rotation occurs if a player rotates from a low-probability shooting position to defend a high probability shooting position. In high school and college levels, this is merely playing the weak side and rotating when necessary. At the NBA level, players are unable to camp in the lane and therefore rotations need to be quicker and smarter.
In this case, let’s suppose Asik rolls to the middle and receives the pass. The smart rotation works this way: Gibson steps up and contends Asik in the lane. Patterson should move to the weak-side block, allowing for a second pass for a lay-up. In this case, Jimmy Butler (#21), who is already hedging at the elbow, slips to deny that pass. This forces either Luol Deng (#9) or Robinson to cover James Harden (#13) and Chandler Parsons (#25), who are both straying at the perimeter. This rotation either results in eating thirteen seconds of shot clock or resulting in a contested shot by Asik.
Illustrating Smart Rotations: Two-Nine and Nail
A way to hedge the offense one-to-two plays ahead on defense is to enforce smart rotations through the Two-Nine and Nail. These regions, illustrated, are given as ellipses in the following image.
The yellow ellipse is a common location for the Two-Nine. Typically, the offensive player will either sag to the short corner on the weak side, or swing into the corner. This forces the Two-Nine to move out of the lane within three seconds.
The blue ellipse is the nail location. This is a more challenging position to cover as this player tends to be guarding the weak-side wing. If this player is, say, Steph Curry or Klay Thompson, then this position may go unfilled. At this point, the opposite corner may pull to the weak side of the lane; intimating a fake nail presence.
So Where Does Mathematics Come In?
Now that we understand the basic concepts of BLUE defense, the question is now, how do we understand how well a team plays BLUE defense? Or more importantly, what physical tendencies do teams, or particular players, have when operating in a BLUE defensive scheme?
We can compute simple calculations such as distance to the rim, distance to ball handler, and location of quadrant on the court. This would give some spatial awareness. More importantly, this gives us awareness of how fast players move.
For instance, let’s consider identifying players closest to desired locations. We can compute the closest players on the weak side of the court to the Two-Nine location and the Nail locations. This would give us a distribution of how likely players are hedging on weak side defense. A really neat exercise was performed in this area (much more was done here than I am leading on; I need to re-emphasis how cool this is…) to show how “ghost players” would react in a defensive situation. Hence the reason for the Raptors’ backdrop for the courts above… Seriously, read the article. Zach did some great reporting there.
We can even take this a step further. In one of my previous posts, I mentioned being in development of a spatio-temporal model that monitors the correlation between players on defense. It builds an empirical orthogonal function computation that tracks the probabilistic movements of players. Combine this with classification of “good” or “bad” rotations; or “scored” versus “unscored” possessions, we have a classification system that readily admits how players reacts along with their reaction times and “ghost” possessions.
A second model we can use simpler models such as convex hulls; work that I had performed back in 2013 for an Eastern Conference team (no longer in use I found out last year). The idea is simple, we mark locations on the court that need to be covered and instill simple rules such as “don’t get beat by your man“ or “do not let penetration into the lane occur”. If the convex hull covers these regions and the ball is safely protected outside of the convex hull, and the offense is located near the boundary of the convex hull, then a team rarely scores. By monitoring the in/out states of the rules set locations noted above we are able to quantify the capability of a team’s defense; or alternatively, quantify an offense’s ability to attack a defense.
Quick Example: Memphis Defense vs. Golden State
Let’s walk through the Convex Hull situation. In an example from the 2015 season between the Memphis Grizzlies and the Golden State Warriors, we find ourselves with yet another sideline screen on the left hand side. The Grizzlies double team Klay Thompson (#11), freeing Draymond Green (#23) for an open jump shot attempt.
The green shaded region is the convex hull of the Memphis defense. Here, we see that no one is protecting the Two-Nine nor the Nail. In these locations, Zach Randolph (#50) shoudl be holding the Two-Nine and Jeff Green (#32) should be hedging the nail. This is due to the fact that Mike Conley Jr. (#11) cannot leave Steph Curry (#30) alone.
While we see the nail is covered, it is not covered as the ball penetrates the convex hull with Draymon Green getting a pass from Klay Thompson. At this point Mike Conley collapses the convex hull by covering Green, leaving Curry wide open for three. However, Zach Randolph bites and turns the hull into a spear, indicating that there are two high probability regions to score: Curry for three (No boundary, left uncovered) and Andre Iguodala (#9) at the low post, uncovered for a lay-up.
This results in an easy lay-up for Iguodala.
Memphis Tries to do the Same!
As we come down the court, Conley immediately gets a sideline screen… yes on the left side but Stephen Curry and Festus Ezeli (#31) BLUE the screen, leaving Randolph free to run the perimeter. Draymon Green sets the nail; thanks to Marc Gasol (#33) passing through the lane. This allows Green to settle into the Two-Nine position while Iguodala hedges to the nail.
Conley kicks the ball back to Randolph and (recall the smart rotations segment above!) Iguodala rotates onto Randolph just enough to stop a field goal attempt or drive, allowing Ezeli to recover. The Memphis gets zero looks at a shot and have to reset the offense.
Capturing the model, we have that Golden States covers all positions and leaves zero high probability shot regions open. This is due to non-movement of Courtney Lee (#5) on Memphis and Gasol’s rotation through the lane, allowing Golden State to rotate smartly and safely.
What Models Would You Build?
Now that you have seen how BLUE/ICE breaks down and where anchor points on the court for smart rotations, how would you develop an algorithm to quantify a team’s capability of playing pick-and-roll defense?