By Chris Baker (@ChrisBakerAM) and Steve Shea (@SteveShea33)
September 23, 2016
If the city of Boston were redesigned from scratch tomorrow, no sane person would weave the confounding knot of one-way roads that tangle the city today. If considered in a bubble, coaching strategies shortly after the introduction of the 3-point line made as much sense as the North End’s roadways after widespread adoption of the motorcar. Of course, professional coaches don’t develop their strategies overnight, nor do players develop their skills so quickly.
NBA offensive and defensive strategies are evolving. They aren’t just changing with the whims of a current coach or executive. They are changing out of necessity. The current NBA talent pool is different from the past. For example, current players are much more accustomed to shooting threes than previous generations that grew up without the shot. There have also been significant defensive rule changes. Hand-checking fouls now make it easier for perimeter players to drive. The abolishment of illegal defense allows teams to collapse into the NBA’s version of a zone (which must respect the current defensive 3-in-the-key).
Current strategies find their roots in past practices. There is value in this model since “survival of the fittest” is at play. The best past strategies have endured the longest (if perhaps with modifications). There are consequences as well. Often elements of past strategies linger beyond their expiration. Ideally, coaches would be able to harvest the wisdom of past experiences without carrying the burden of a bias towards past success when designing strategies that consider the constraints of today’s game. But that isn’t realistic.
Analytics isn’t (usually) biased towards past tendency, and sometimes analyzing the game can feel like taking a snapshot at the Target checkout just after a new register has been opened. We wonder why Phil Jackson is standing 10 deep at aisle 7 when Daryl Morey is alone in aisle 2. (Sam Hinkie is first at an unmanned register in aisle 6 playing the odds that the store will open it soon. If Joel Embiid exits the break room in the next 30 seconds, he’ll look like a genius.)
Red Auerbach once proclaimed, “Basketball is like war in that offensive weapons are developed first, and it always takes a while for the defense to catch up.” Recent years agree with Red; NBA offenses are evolving faster than defenses are responding.
The components of the modern offensive evolution can be subtle, such as replacing a few pick-and-rolls with dribble handoffs or encouraging less pursuit of offensive rebounds to better defend transition.
There is no subtlety to the growing importance of the 3-point shot. It’s become the hallmark of today’s game. In 2015-16, NBA teams averaged 24.1 3-point attempts per game. That’s up 33% from 18.0 attempts per game just 5 years earlier, and up nearly 90% of the 12.7 attempts per game in 1997-98 (which was the first season after a 3-year trial of a shorter 3-point line).
The 3-point growth shows no sign of slowing down. Teams shot 1.7 more 3-point attempts per game last season than the prior season (although a faster pace league-wide contributed to that growth).
It’s not that weaker teams are heaving more threes in desperation. The Eastern and Western Conference Champions, the Cavs and Warriors, were in the top three in both 3-point attempts and in % of the team’s FGA from behind the 3-point line. 10 of the top 11 teams in 3-point attempts made the playoffs.
The increased usage of 3-pointers is justified. As the NBA talent pool became increasingly efficient from deep, the shot’s value surpassed almost any FGA besides a wide-open dunk. In 1982-83, NBA teams shot 23.8% on 3-pointers. That’s equivalent to 0.71 points per FGA. By comparison, teams shot 49.2% on twos or generated about 0.98 points per shot. A two-pointer was a much better shot. Today, teams are still generating about 0.98 points per two-pointer, but improved 3-point shooting has teams averaging 1.06 points per 3-point attempt. For the Golden State Warriors, 3-point attempts generate 1.25 points per shot. (That’s why they take 31.6 of them a game.)
When we focus on the “good” 3-point attempts, efficiencies improve significantly. Teams averaged 1.13 points per corner 3 attempt and 1.11 points per catch-and-shoot 3-point attempt last season.
Compared to a “good” 3-pointer, a “bad” two-pointer looks inexcusable. Teams scored just under 0.8 points per mid-range jumper in 2015-16. An average team could pick up 2 points per game just by replacing 6 mid-range jumpers with 6 corner 3s.
The threat of a 3-point shot can be as valuable as the shot itself in that it provides spacing for the rest of the offense—room to drive, cut or post-up. For more on this idea, see our previous post.
We all agree, but have yet to see
Of course, not all NBA teams adopted 3-point usage at the same rate or embraced the improved spacing provided by 3-point threats to the same degree. Aisle 2 in the checkout line has been open for a while, but we’re still far from equilibrium (as the Bucks so kindly demonstrated last season). The NBA teams that were late to the 3-point party still must properly configure their rosters and transition their offensive schemes to come close to the Cavs or Warriors in effective 3-point usage.
But on-the-court equilibrium isn’t realistic. There will always be talent disparities. There will always be teams at various stages of competitiveness—some bringing back a core for their 10th consecutive playoff run while others are reloading with young talent. The true measure of where we stand in the offensive evolution lies in where basketball executives and coaches side on the issue.
It’s now safe to say that all NBA teams are onboard the 3-point train, even if strategies differ as to how best to implement the shot. If there were still any doubt after the 2015-16 season, the last great hurdle was crossed when Knicks’ President Phil Jackson consented, “The 3-point line has become our affection, because it means more when we make a 3-point shot.” (Thanks Phil.)
An opposite (but not equal) reaction
If offenses have changed so radically, so then must the defense. As the 3-point shot became more efficient and as teams began implementing offenses to specifically generate 3-point attempts, teams needed to devote more defensive resources to paroling the 3-point line, running off shooters and heavily contesting shots. The defensive changes needed to be as extreme as the offensive shift.
As Red Auerbach realized, the defensive adjustments always lag behind the offensive maneuvers. The NBA is in the early stages of a defensive 3-point revolution, a counter to the offensive development. Here, the Target checkout phenomenon is real; some teams are more progressive than others, not just in on-the-court performance, but also in front office philosophies.
Penetration and Kick outs
If an NBA player wants to pull-up for 3 in traffic, there is little that the defense can do about it. Fortunately for the defense, that’s not the type of 3-point attempt that they need to be concerned about. Rather, it’s the catch-and-shoot opportunities that are troubling.
NBA defenses won’t intentionally leave a capable shooter open enough to catch and shoot a 3 in rhythm. Instead, NBA offenses have to penetrate, draw help defenders and force defensive rotations to create enough space for perimeter shooters.
Understanding how teams defend the 3 means understanding how teams defend penetration and kick outs.
The best defense of a catch-and-shoot 3 is to not allow the shot. In theory, there are two ways this can happen. First, a team can guard the perimeter players so closely that a kick out isn’t attempted, or when it is, the perimeter player is coaxed to drive or pass. The second method would be to position the defense to create turnovers either by occupying passing lanes or by aggressively trapping the ball handler before the pass attempt.
That’s the theory, but are teams practicing either strategy and if so, are they successful? There is a remarkable amount of real estate on the 3-point line, and many teams now play lineups with at least 3 perimeter threats. Can teams consistently and significantly create turnovers or reduce catch and shoot 3-point attempts?
Yes, and the defensive systems that are being employed are remarkably different than those from the past.
The minimal help model
All other variables the same, the closer the shot, the easier it is to make. Prior to the 3-point line and even in the first few years of its existence, the best shots (by far) were those near the hoop. As a result, NBA defenses protected that region at all costs. Teams collapsed with help defense to the best of their ability (under the old illegal defense rules). In those days, forcing a kick out on penetration was a win for the defense.
Today, the best defenses are doing the opposite. They are sending minimal help on the drive. In particular, teams are not leaving the corners open, and they are terrified to leave an elite shooter (e.g. Steph Curry, Kyle Korver or J.J. Redick) alone anywhere behind the arc.
To assess defensive strategy on penetration we’ll use measurements of offensive and defensive stretch, which were introduced in Basketball Analytics: Spatial Tracking.
Consider the image below taken from an NBA.com movement animation. It is the 3rd quarter of the Clippers-Warriors game on March 31, 2015. Notice the blue polygon wrapped around the Clippers’ defenders and the red polygon wrapped around the offensive players. The area of the blue polygon is CHAD (the Convex Hull Area of the Defense), and the area of the red polygon corresponds to what we called the Convex Hull Area of the Offense (or CHAO).
When CHAD is close to CHAO, we have a situation like the one above where the defense is aggressively guarding the 3-point line.
The second image below was taken from a Knicks-Clippers game on March 25th. Here, the Knicks are well spaced. Their area (or CHAO) is large. However, the Clippers have a very small CHAD. When CHAD is small relative to CHAO, the defense has collapsed and left open players on the perimeter.
We looked at every halfcourt possession in the NBA in 2014-15. (In other words, we eliminated transition.) In those halfcourt possessions, we marked the first instance the offense penetrated (moved from possession outside 15 feet to possession inside 10 feet.) This could be a pass to the post or a cutter, or it could be a drive from the perimeter. Using spatial tracking coordinates provided by SportVU, we looked at the position of all players on the court the first instant the offense had possession within 10 feet of the hoop. Then, we calculated the offense’s spacing (CHAO) and the defense’s stretch (CHAD). The teams with the smallest difference in CHAO and CHAD are the ones that help off the perimeter the least. Here are the results
|Rank||TeamCode||CHAO-CHAD (sq. ft)|
Chicago, Portland, Cleveland, San Antonio, and Golden State helped the least. These teams were the leaders in the minimal help model, a defensive strategy on penetration that is in direct contrast to traditional defense.
(We don’t mean to suggest that the above 4 teams employ identical defenses. The teams employ defenses strategy that align in their stretch during penetration and in their strategy to reduce 3s on kick outs.)
Does it work? There are costs to not helping as much on defense. It means less obstacles for the penetrating player, and possibly higher opponents’ efficiency around the rim. It can mean less resources around the hoop for rebounds. So, if a team is going to intentionally not help, there must be a benefit. The goal of not helping is to prevent catch-and-shoot threes. Were these teams able to do that?
We looked at how many catch and shoot 3-point attempts (C&S 3PA) each team gave up in 2014-15. We adjusted for opponent tendencies by looking at each game individually and recording how many more or less C&S 3PA a team allowed than their opponent usually attempted. For example, if San Antonio held Golden State to 15 C&S 3PA, which was 6 less than they usually attempted, it was seen as a reduction. In contrast, if San Antonio allowed 15 C&S 3PA from Minnesota, which was about 5 more than they usually got, it was seen as an increase. We adjusted for pace by looking at percentages of typical opponent “shots” (FGA+0.44FTA) instead of totals. For ease of interpretation, know that 1% equates to approximately one shot per game.
The following chart plots CHAO-CHAD to the above-described percentage. There is a strong correlation. The teams that help the least (have the smallest CHAO-CHAD) are able to reduce opponents’ C&S 3PA. The leaders in this category (San Antonio) are holding opponents to 3-4 less C&S opportunities per game than they typically get. That’s a sizable chunk when teams are averaging 16.5 of these shots a game.
Preventing opponent C&S 3-point attempts has lessened opponents’ shooting efficiency. The top three defenses in opponent points per shot in 2014-15 were Golden State (0.940), Chicago (0.946) and Portland (0.952). San Antonio was a healthy 6th at 0.968. Those were the top 4 in terms of lowering opponent C&S 3PA.
The swarming defense
With 4:20 left in the 1st quarter of a January 31st, 2015 matchup between Portland and Milwaukee, Nicolas Batum feeds Lamarcus Aldridge on the block. Upon the penetration, all five Bucks sag into or around the paint. (This wouldn’t have been allowed before the NBA replaced it’s illegal defense with the defensive 3-in-the-key.) This leaves two Portland players (including Damian Lillard) open above the break. Portland is helping Milwaukee by having no players available in the corners.
Aldridge turns towards the paint, but is immediately met by Lillard’s man, Brandon Knight. Knight knocks the ball loose for a turnover.
Knight had to leave Lillard open for a C&S 3 when he trapped Aldridge. It was a risk. Most times, Knight isn’t going to get the steal. NBA players tend to see traps coming, and are poised and strong with the ball. A full court press might work well against high school players, but would be fruitless against he ball-handling and passing ability of the professionals.
Knight’s gamble wasn’t a full court press and it doesn’t have to work every time. Milwaukee’s swarming defense understood that it would give up a good amount of C&S threes, but believed that they would get enough turnovers to be an efficient defense overall. They were correct; Milwaukee had the 3rd best defensive rating that season (per Basketball-Reference.com).
When the minimal help model prevents a 3, it’s usually exchanging that 3-point attempt for a different (and hopefully less efficient shot). When the swarming defense prevents a 3, it’s through a turnover. If the swarming defense can keep its turnover % high enough, it can offset the increased efficiency realized by the opposing offense through more C&S threes.
In 2014-15, Milwaukee, Atlanta and Philadelphia all saw at least some degree of defensive success with a version of this swarming defense. Their success is reflected in the percent of turnovers they induced. All three teams forced opponents to turnover the ball on at least 14.9% of possessions. They were the top 3 teams in this category.
Miami also employed a swarming defense. After being 1st in the league in opponent turnover % in 2013-14 (at 15.8%), Miami dropped to 8th in 2014-15 (with 14.2%). The inability to turn the swarming gamble into a turnover often enough meant the team slid to 21st in defensive rating.
If swarming defenses are consistently collapsing on penetration, we should see that in the spatial tracking data. The table of CHAO-CHAD data is reproduced below. The four teams that collapsed the most were Milwaukee, Miami, Atlanta, and Philadelphia.
|Rank||TeamCode||CHAO-CHAD (sq. ft)|
We suggested that swarming defenses will give up more C&S opportunities. Recall the scatter plot that mapped defensive stretch on penetration to percent influence on opponent C&S 3PA. The teams that collapsed the most (produced the highest CHAO-CHAD) also gave up more C&S 3PA.
Many of the C&S opportunities that swarming defenses give up come from the very efficient corner. 31.5% of Bucks’ opponents’ 3PA came from the corner. It was 28.8% from Atlanta’s opponents. Those were the two highest percentages on the season. In contrast, only 20.1% of Chicago’s opponents’ 3PA came from the corner. That was the lowest % in the league.
We discussed earlier how we still see great variation in perimeter shooting ability and usage among NBA offenses. The swarming defense would appear to be ideal against a team with minimal perimeter shooting since when the kick out to the open shooter is successful, the shooter will be less efficient on the shot. In contrast, it would seem that a swarming defense would struggle against a great passing and perimeter shooting team like Golden State.
The ideal defense might be one that can employ both defensive strategies. However, the 82-game regular season provides little opportunity for teams to prepare for specific opponents. Often teams won’t have a real practice between games.
Since the analysis averages CHAO-CHAD across the season, a team that alternated styles could appear as central and non-distinct. We did study the game-by-game numbers, and as the difficulty of this strategy would suggest, no team altered strategies in a way that correlated with the 3-point shooting of the opponent. Yes, teams do scheme for particular 3-point threats, such as Curry or Korver, but otherwise swarming defenses swarm. Any adjustments for individual players were not significant enough to alter team stretch totals.
The defenses on the extreme in CHAO-CHAD outperformed those in the middle. The bottom 8 teams in defensive rating were ranked between 13 and 26 in CHAO-CHAD (where the minimal difference was ranked 1st). The skew here suggests that not collapsing is generally better than swarming. The minimal help model also wins when we look at the top of the defensive rating board.
For the sake of this argument, suppose that a CHAO-CHAD < 285 indicates a minimal help model. No minimal help model finished in the bottom 12 in defensive rating.
There is a difference between system and execution. The minimal help model will only be successful if it actually reduces opponents’ C&S 3PA%. Consider the model effective if it reduced opponents C&S 3PA by at least 1% (of the total offense). There were 5 effective minimal help models (out of the 6), and those 5 teams were in the top 11 in defensive rating. Two of these models (GSW and SAS) were the top 2 in defensive rating.
If we consider a team as swarming when CHAO-CHAD>300, we have 5 swarming models. This model is effective only if the team is able to induce a high amount of turnovers. Let’s make that cutoff 14.5%. We then have 3 effective swarming models. Two of them (Milwaukee and Atlanta) were in the top 6 in defensive rating, while the third (Philadelphia) was 13th.
The bottom 17 teams in defensive rating qualified as neither an effective swarming nor an effective minimal help model. In other words, all 8 effective modern defensive models were in the top 13 in defensive rating.
What actions should a team take now with the above information in hand?
This article is about defensive strategy, but we can’t help but again suggest that modern offenses need to be a threat both inside and out. Lineups need at least one player that is dangerous around the hoop. With modern hand-checking fouls and the typically superior free-throw shooting of perimeter players, this is often a guard that can drive (perhaps off a screen).
During penetration, offenses need to force defenses to make tough decisions. A player that is efficient attacking the hoop begs help defenders to collapse. Spacing the floor with 3-point threats (and we recommend at least 3) makes it dangerous to leave the corner unmanned.
We suspect that as offenses get better at shooting 3s, teams will help less on penetration. In other words, the minimal help model will be become the most popular. Thinking ahead, what does this mean for offenses? It’s possible that this opens the door for the return of the dominant post center (in the mold of Olajuwon or Shaq). More likely, NBA offenses will be able to counter with stronger and more athletic driving ball handlers that won’t be as affected by one defender on their shoulder. LeBron is the ideal, but this might also be Ben Simmons in Philadelphia or Giannis in Milwaukee (as examples).
Teams need to decide on a defensive strategy. Ideally, teams would have the flexibility to play both modern models described above. However, it’s not realistic to expect young players to be able to switch from one helping extreme to the other on the fly. The middle ground defensively, which can happen through indecision, hesitation and confusion, is the worst. Thus, it probably makes sense for teams (and especially younger teams) to commit to predominantly one style for the regular season.
As teams trend towards better perimeter shooting, we suspect that the minimal help model will surpass the swarming defense in effectiveness. The swarm will have it’s role, but in small doses like a blitz in football.
Length and athleticism in defenders is remarkably helpful regardless of system. A perimeter shot contested by Kawhi Leonard is different than a perimeter shot contested by Jason Terry. And if the perimeter defender is left with little help when his player penetrates, a paint shot contested by Leonard is different than a paint shot contested by Terry.
In addition, length and athleticism translates to positional versatility. It allows players to switch screens without creating major mismatches in speed or size. Switching screens cuts off the space that offensive players use to get a step to the hoop or launch a 3.
We scan the NBA landscape and see elite offenses with 3-point shooting at their core. The natural reaction is to expect NBA defenses to be designed with preventing the 3 as a core principle.
Certainly some teams have adapted quickly. San Antonio and Golden State both employ effective minimal help models (and not surprisingly, are very successful franchises). Coach Tom Thibodeau pioneered the model as an assistant coach for the 2008 Champion Celtics. He then employed his defensive model with great success for years in Chicago.
Milwaukee and Atlanta have found defensive success with a modern swarming model. Their success is in part due to a focus on bringing in long and positionally-versatile wings that can switch screens and occupy passing lanes.
Yet, we still see a number of teams seemingly unsure of what to do in response to the 3-point revolution. To understand why teams appear so stubborn, we have to understand where today’s coaches and managers came from.
Many of the executives and coaches in today’s NBA have been involved in high-level basketball for 30 years or more. The first 25 of these years, these individuals never encountered a team like Golden State. Furthermore, these coaches and managers are where they are because they were so successful in the past. We have individuals that have seen decades of success with certain systems and philosophies. Why would we expect them to change so quickly?
We can’t ignore the practical challenges of finding the right personnel for a modern defensive system. When the 3-point shooting giants first presented, it was also new for NBA players. A minimal help model might be nice in theory, but how successful would a team be at implementing it if all of its players have no experience in anything similar?
Changing defensive systems can also require a significant revision to the roster. Teams might have players under contract for several years whose value would diminish tremendously if the team changed defensive strategy.
While we sympathize with the challenges NBA decision makers face when trying to counter the 3-point revolution, the challenges do not negate the reality that teams must adapt.