Basketball is a sneakily complex game, and the naked eye cannot always ascertain the full story.
A player might appear dominant, but is in fact taking less-efficient shots than his teammates would be taking were he not on the floor. A center may be blocking shots left and right, but is doing so only by overplaying, leaving his man open for several dunks and putbacks.
This is why advanced statistical analysis is not only useful, but necessary in the modern NBA. With an ultra-competitive league and a constricting salary cap, throwing big contracts at volume scorers and shot-blockers can cripple a franchise.
But while advanced metrics and stats have revealed hidden truths about basketball, they have also clouded the game. Our ability to watch and use common sense has been compromised by the notion that the numbers describe a reality that's truer than what pure observation is capable of telling us.
Take, for instance, the isolation play.
The play was once thought of as the best in basketball. Get the ball to your best player, give him space and let his man try and stop him from scoring.
Of course this kind of approach has always been viewed as limited. Having your best player possess the ball too frequently leads to fatigue on his part, and running plays designed to free up a man for an easy layup or an open catch-and-shoot opportunity has always been thought of as a pretty efficient way to score.
But when the momentum is swinging to your opponent, or your best guy has a mismatch, or he's starting to feel it, or the game is winding down...going to an isolation set has always been seen as a great option.
Recently, this has changed. Advanced statistics have suggested—or people suggest they have suggested (more on that later)—that isolation sets are inefficient. Looking at Synergy Sports' points per possession data, Henry Abbot wrote the following in an Aug 2012 column for ESPN the Magazine:
Plays involving off-the-ball cuts (1.18 points per possession) and transition plays (1.12 ppp) are by far the most efficient, followed by putbacks (1.04 ppp) and pick-and-rolls in which the ball reaches the hands of the rolling man (0.97 ppp). And the least efficient? Isolation plays, good for only 0.78 points per possession.
Synergy's numbers are no doubt accurate. The problem is that the conclusion that Abbot and so many others have drawn from these numbers is not.
The first issue with using points-per-possession data to condemn isolation is that not all possessions are the same.
Let's look at the key differences between each type of play that Abbot refers to.
- Plays involving off-the-ball cuts are highly effective because the cut is typically made as a reaction to a great screen, a defensive lapse or a double-team elsewhere. The pass being made to the cutting player is usually made only if there is an open lane to throw the ball through. Finally, these plays are drawn up by all-time great coaches like Gregg Popovich and Phil Jackson a disproportionate amount of the time.
- Transition plays are incredibly efficient because all five defenders rarely get into position quickly enough to stop them. The offense often has a man advantage and always has more open space than it would in a half-court set.
- Putbacks first require a shot to missed and an offensive rebound to be grabbed. Once this has happened, the offense usually has exceptional position down low, and the defense typically is scattered.
- Pick-and-rolls toward the basket are effective plays, but several things must happen to create this situation. The pick must be set, the trailing defender must be slowed, the screener's man must switch and the screener must quickly recognize this and roll toward the hoop. The ball-handler must then react and make an accurate, catchable pass.
- Isolation plays require the ball to be possessed by one offensive player and space to be cleared by the other four.
It is strikingly clear how flawed this data is. The four plays besides isolation all require several things to break right before even being considered "plays." Isolation, meanwhile, can be set up at any time. It requires no turnover, no offensive rebound, no precision pass and no great X's and O's coach.
As a result, it is an inherently less efficient play. But alas, a team cannot decide to run a transition play. A coach cannot draw up a putback. An offense cannot genuinely choose between running an iso or running a pick-and-roll that is automatically successfully executed.
If putbacks are considered plays that are worth discussing in terms of points per possession, then we should not talk about isolations, but rather isolations in which the ball-handler gets by his man and makes it to the rim.
If transition opportunities are considered plays, then an isolation in which the defender gets crossed up and hits the floor should also be considered a "play."
If pick-and-rolls only count as plays when the ball makes it to the roller's hands, then isolations should only count as plays when they involve a 6'2" point guard who is switched onto an All-Star small forward.
This is not to say that the isolation is the only play that can be run without a pre-existing advantage. Pick-and-rolls are still plays even if they are not successful. Teams can swing the ball and take catch-and-shoot threes. A point guard can drive and dish. There are several ways to run a half-court offense.
Even so, isolation and the idea behind it is instrumental in the majority of successful, efficient half-court sets.
The Threat of Isolation
During Game 2 of the Oklahoma City Thunder's series with the Memphis Grizzlies last May, catch-and-shoot specialist Derek Fisher went off. Would anyone call Fisher's second-quarter explosion the result of isolation?
Advanced stat aficionados would not. They would call it a result of ball movement and corner threes. Isolation, they believe, is never to be connected to anything efficient, good or commendable.
But without the threat of isolation, those wide-open looks would have never existed.
Kevin Durant is one of the toughest players to guard in the NBA. He, along with LeBron James, Carmelo Anthony, James Harden and several other elite scorers require far more attention than defenses want to give them.
Ideally, defenses want to be able to guard everyone straight up. This allows defenders to never have to rotate, never have to leave shooters and therefore minimize offensive rebounds. The more simplified a defensive approach is the better—but only when it's possible to succeed with this simplicity.
With a player like Durant on the court, simplicity leads to disastrous results. Putting a single defender on Durant will cause the Thunder to get him the ball right away. If help doesn't come, his teammates will clear out, isolate Durant and let him destroy his defender. Once his defender is beaten, if there is no rotation or help, Durant will take the ball to the rim and finish.
This is the most efficient play in basketball. The reason the numbers do not back this up is that this hardly ever happens.
Defenses do not allow Durant or Durant-level players to work one-on-one against defenders. They bring double-teams, forcing a pass and thus stopping an isolation from ever occurring. When they do not double-team, they will always have a second line of defense ready to rotate over, knowing that the guy on Durant is liable to be demolished.
Not only do defenses know better than to try and stop elite scorers with one man and no help, but the coaches of these stars know that the defense knows this.
As a result, a coach like Scott Brooks will rarely run a straight isolation play for a guy like Durant all the way through. Rather, they will get the ball to Durant, have him assess the defense and react accordingly.
The video above demonstrates many of the possibilities here. In one instance, Durant is double-teamed. He quickly gives the ball up to Thabo Sefolosha, who swings it to Fisher for a corner three. In another, Durant beats his man, starts his drive, senses the defense collapsing and kicks it to Fisher for another corner triple.
Near the end of the clip, the Thunder attempt to isolate Durant on the right side of the court. The defense refuses to allow it, keeping two defenders on that side. Durant still breaks down his man and gets inside, only to kick it to Kevin Martin in the corner. This causes another rotation, which leaves Fisher open for a three.
Only the final play saw an isolation attempted to be carried out—and even then it did not play out completely. Still, all three baskets were the result of the unwillingness of the Memphis defense to allow Durant to attack one-on-one.
This cannot be shown with an advanced statistic; it can only be seen through observation.
Before denouncing advanced stats for failing to recognize how some facets the game is actually played, it should be noted that the stats are not really at fault. It's the interpretation of the numbers that is far more worthy of blame.
As with all statistics—sports related or not—the same numbers can be framed in different ways. In the case of advanced basketball metrics, the analysis often stops just short of reaching the point of disproving itself.
For example, a big deal is made over the greater efficiency of the three-point shot compared to the two-point shot, but the fact that shooting more twos results in more free throws and foul trouble is ignored. The fact that threes tire out the opposition less than pounding it in the paint is not considered.
The "advanced" data that people say reveals Kyle Korver to be more efficient (.641 effective field-goal percentage) than Zach Randolph (.457 eFG%) is in reality only saying he earns more points per field-goal attempt. It does not attempt to measure his true efficiency, because there is no way to measure the impact of Randolph's relentless elbows to the opposing big men. There is no way to measure the energy conserved by Korver's defenders as he cuts and shoots, rarely driving or posting up.
The same ignorance occurs with isolation criticism. Let's pretend for a minute that Synergy's points per possession data is truly based on a level playing field, even though we proved it was not. We are assuming that every play is equal, and that there is no value to the threat of isolation other than the isolations that actually come to fruition.
Imagining this reality, the isolation is still a more effective play than it is given credit for.
While isolations may result in fewer points per possession, they also result in fewer turnovers. A missed shot is usually preferable to a live-ball turnover that oftentimes fuels a fast break (a transition play). Moreover, a missed shot at least creates an opportunity for an offensive rebound (and a possible putback).
The isolation is also a highly effective game-management tool. A coach can isolate an opposing defender who is beating his team on the other end. By doing this, he can make that player work harder and expend energy. If he can switch to get them mismatched, he can get that player into foul trouble.
When used right, isolation can change the entire makeup of a game. If a team is being beaten in an uptempo, track-meet style game with transition dunks and open three-pointers, going to an isolation offense will slow down the pace. Even if the opposition has the personnel to guard any isolated player one-on-one, an iso approach can turn a frantic game into a plodding one, thus destroying the rhythm of an uptempo team.
The statistics do not show this. Moreover, they do not attempt to show this. The problem results from fans and other observers misreading these statistics and taking them to be profound, all-inclusive numbers and meaning-of-life-revealing truths rather than specific data that shows us only one small side of an endlessly complex game.
The Future of Isolation Basketball
A deeper look—or, alternatively, an emergence back out of the depths, where the light is easier to see—shows that the isolation play is an integral part of winning basketball, as it always has been. The question is, will it always be?
Advanced statistics emerged in basketball mainly as an attempt to follow in baseball's footsteps.
From the introduction of sabermetrics in the 1980s to its initial acceptance in the late 1990s and to its current universality, advanced stats have completely changed baseball inside and out.
Batting average and home runs were challenged by on-base percentage and slugging percentage. Then those were replaced by weighted on-base percentage and weighted runs created plus. On-base and slugging percentages were essentially fancier compositions of basic stats, whereas the latter numbers actually quantified and qualified uncharted elements of the game.
Similarly, basketball analysis has improved since the creation of metrics like offensive rating, win shares, effective field-goal percentage and opponent's player efficiency rating. However, the game is only beginning to take the next step into the truly revolutionary realm.
Player tracking data is opening new possibilities surrounding basketball analysis. In Kirk Goldsberry's Feb 2014 feature on Grtantland.com, he writes about expected possession value, or the idea that every moment in a basketball game is a weighted situation:
Every “state” of a basketball possession has a value. This value is based on the probability of a made basket occurring, and is equal to the total number of expected points that will result from that possession. While the average NBA possession is worth close to one point, that exact value of expected points fluctuates moment to moment, and these fluctuations depend on what’s happening on the floor.
...For example, imagine LeBron James holding the basketball completely unguarded underneath the basket. We would expect him to score two points. The EPV at that moment would be very close to two. Conversely, imagine Dwight Howard holding the ball 40 feet from the hoop with one second remaining on the shot clock and three defenders in his face. It’s highly unlikely that Howard would score. That moment would be ascribed an EPV very close to zero.
Goldsberry goes on to discuss the infancy of this type of statistic, and that it may never emerge as a basis for the future of basketball analysis. Regardless, the fate of the isolation play will not be determined by how standard this analysis becomes. A look back to baseball proves that.
Even given where the MLB is along its advanced statistical timeline, old ideas of winning baseball have endured. The stolen base and sacrifice bunt remain prevalent, as do the concepts of clutch hitting and pitching to contact.
The reason for this is simple: No statistic can compete with the power of observation.
This is not because we want to believe what we see; there is too much at stake with modern professional sports for us to be willingly deceived. Rather, it is because observation yields information about a player's value that statistics still cannot and will likely never be able to deliver.
Coaches understand this. Players understand this. The people who are the closest to the game may not have the best understanding of advanced metrics, but they have an innate sense of the sport that is just as rooted in what's actually happening as these supposedly all-revealing statistics are.
So, to answer the question of why isolation basketball still exists is pretty easy: It works; defenses know it works, and its impact on the game is much deeper than the analysis that attempts to tear it down.
Stats courtesy of Basketball-Reference.com unless otherwise noted.