Stephen Curry and Guarding the Unguardable

May 15, 2015; Memphis, TN, USA; Golden State Warriors guard Stephen Curry (30) shoots a three point shot in the second half defended by Memphis Grizzlies guard Mike Conley (11) in game six of the second round of the NBA Playoffs at FedExForum. Warriors defeated the Grizzlies 108-95. Mandatory Credit: Nelson Chenault-USA TODAY Sports

If Stephen Curry’s MVP season can be traced back to one skill, it is his ability to take and make three pointers he creates on his own, and to do so at greater volume and efficiency than anyone else in the league.

The notion of “getting one’s own shot” is s tricky one. Often times what is meant is the ability (and, as importantly, willingness) to shoot jump shots off the dribble. Often left unsaid is that creating one’s own shot is usually creating a bad shot. Pull-up jumpers[1. As defined by SportVU — shots taken after one or more dribbles from more than 10 feet.] had an overall accuracy of 38.7% in the NBA this season. Even factoring in offensive rebounds, an offense of only these shots would be the worst in the league by a big margin.

Still, creating shots is an important skill — some players (like Chris Paul) are good enough at these shots that they must be guarded closely, opening up more efficient avenues such as drives to the basket or opening up teammates for more valuable catch-and-shoot possessions[2. Catch-and-shoot jumpers had an effective field goal percentage (eFG%) of just under 50% on the season, per SportVU.]. In general, these “playmade” shots are far higher efficiency for an offense than the alternative. Assisted shots are “worth” almost 1/3 of a point more per shot than unassisted attempts. The basis of playmaking is drawing one or more extra defenders to the player with the ball. And this is where the value of players who really can “get their own” with some degree of proficiency show their value.

Defining this ability statistically remains problematic. Percentage of made field goals coming via assist is a reasonable first pass, though an incredibly noisy one, with factors such as role and shooting location[3. 52.2% of two pointers were assisted in 2014/15, whereas 84.4% of made threes came directly from a pass.] playing as large if not larger factors in this number than a player’s actual creative ability. To really measure how much better an individual performs when set up by a teammate or when foraging on his own, we need to know not just the number of makes, but attempts. While SportVU has the promise of this sort of data, it isn’t yet publicly available. We know who is setting up shots, but not where these shots are occurring or who is taking them if they aren’t made[4. Thus generating an assist.].

However, getting back to Curry, there are a certain selection of shots about which we can at least make educated guesses in terms of how many

attempts

(not just makes) were assisted. For the most part, threes are fairly homogenous shot attempts—a player creates space with some semblance of shooting rhythm, elevates and fires. Most of the time, this is off of a catch-and-shoot[4. 77.6% of all three pointers were taken with no dribbles between catch and shot.]. The overwhelming majority of threes with no dribble are assisted shots, 98.1% to be exact this past season. Unsurprisingly, once the player puts the ball on the deck, shooting percentages drop as do the proportion of makes coming via assists:

As the chart shows, a good percentage of threes taken off of one or even two dribbles generate assists. These are mostly in transition or of the “Steph Step” variety, meaning a kick out to a mostly open player who pump fakes the defender into flying past before taking a rhythm dribble and stroking it:

Even though the shot above is off the dribble, it has more of resemblance to a catch-and-shoot than Curry really “creating” his own opportunity. Something has happened in the flow of the offense to allow him to receive the ball with an advantage on his defender, which he exploits ruthlessly.

In order to truly capture, or at least estimate, the skill level of individuals in terms of creating three point looks against a more set defender, plays such as the above should probably not be included. Without going through all 56,000-plus three pointers taken in 2014/15 manually, it’s impossible to get an exact count. But we have a rough idea. About 3/4ths of all threes made off of one dribble and a third of those using two dribbles were assisted[5. The proportion “potentially” assisted is probably slightly lower, as shots truly set up are very likely made at a higher clip than those not coming via the work of a teammate.].

By looking at the number of attempts taken off of zero, one and two dribbles[3. I’ve chosen to ignore shots taken off of three or more dribbles both for sake of simplicity and the fact that an assist awarded after that much dribbling is questionable and almost certainly would not be scored as a “potential” assist via SportVU. I’ve also chosen to use only shots taken from 30 feet or closer to weed out desperation heaves.] it is possible to estimate the number of “assisted attempts” for each player. Further, since the number of assisted and unassisted made threes is known, it is also possible to approximate the percentage each player makes on assisted versus unassisted threes.

The following chart captures some of the characteristics of some of the leading long-range shooters in the game in terms of both proportion and accuracy of assisted vs. unassisted threes:

First of all, Curry probably led the league in unassisted 3FGA. I say probably because there are reasons to believe other high attempt players like Harden and Lillard who appear to have shot better on unassisted threes were probably taking a harder than average mix of one and two dribble three-pointers[6. Other than Curry, most of the other high attempt, high estimated percentage on unassisted shots such as Harden, James and Lowry were actually assisted a lower number of makes than their mix of shots initially suggests. In other words, it seems likely more of Harden’s misses off of 1 or 2 dribbles were “created” shots than the league average would suggest, meaning in actuality he shot somewhat better on assisted shots and somewhat worse on unassisted shots.]. Secondly, Steph Curry is ridiculous. While NBA average on unassisted threes is probably around 30%, Curry shot around 43%! On unassisted shots only he would have been 6th in the NBA in 3PT%.

Combining the volume and accuracy, it’s clear why this is such a singular and devastating skill. A self-created three is a shot defenses will live with from most NBA players. From Curry, that shot alone would have been the best offense in the league. All of a sudden, extra help is needed in a situation where it is unnecessary against 29 other teams.

Getting back to the beginning of the article, this extra defensive attention to Curry is a major reason why Golden State had one of the most cooperative offenses in the league, with 58.4% of their field goal attempts coming via potential assist[10. This was second only to Atlanta’s 60.6%. Houston was also slightly above average, assisting on 54.3% of field goal attempts, leaving only Cleveland as below average among the four conference finalists. And the Cavs feature both 2 of the best individual isolation scorers in the league and a reliance on offensive rebounds, both of which tend to reduce assist possibilities.]. The Warriors had a 62.8% eFG% on their assisted shot attempts, best in the league. Yet even that figure is preferable to letting Curry “get his own” from three, as his 43% accuracy translates in 64.5% eFG. In effect, Curry’s shooting offers a defense a red pill and a blue pill, except both are poisonous.