Nylon Calculus: The evolution of Stephen Curry
Stephen Curry is “not a point guard.” Not a “pure point guard”, anyways. He’s just a “shooting guard playing point.” Stop me if you’ve heard this one before.
The origins of this urban legend are clear enough. Curry was a converted point guard at Davidson. Scouts pegged him as a guy who was “probably never going to be a pure playmaker,” a “capable facilitator” at best — more Mo Williams than Steve Nash. He wasn’t considered “a natural point guard that an NBA team [could] rely on to run a team.”
Understandably then, at the NBA combine, Curry was listed as “PG-SG” along with fellow hybrid Tyreke Evans. Even the Warriors were initially unsure of Curry’s role. Despite having used the No. 7 pick to get him, former GM Larry Riley emphasized after the draft that he did NOT necessarily consider Curry the “point guard of the future” for Golden State.
Bob Myers eventually clarified the situation by drafting (definitely a shooting guard) Klay Thompson in 2011, trading away the (harder to label guard) Monta Ellis later that season, and solidifying the Splash Brothers as the starting backcourt the following year.
Despite the fact that Curry was unambiguously his team’s point guard by the start of the 2013-14 season, opposing GMs paid him the backhanded compliment of naming him the third-best SHOOTING guard in the league. The GMs repeated the indignity the following year.
Any doubts about Curry’s point-guard legitimacy are surely “old news” though, right? The guy was the freaking unanimous MVP, after all. What more vindication could he possibly need?
It’s true, Curry has recently received plenty of the accolades and recognition he is due; yet, radio hosts continue to accuse Curry of being a “system” point guard. Legendary ex-players still deride his perceived shortcomings: “not a great playmaker,” “not a dominant player.” And the critics keep minimizing Curry’s contributions by characterizing him as a superfluous “decoration” in Kevin Durant’s dream house, nothing more than a “collaborative talent”.
There are real reasons why smart people misconstrue Curry’s role on the Warriors. He DOES play off-the-ball more than many of his point-guard counterparts in the NBA. But, the quintessential responsibility of a point guard is to help his teammates score, and Curry is doing THAT better than anybody else in the league this year.
Curry, no longer a floor general
We have expectations of a traditional point guard. He is meant to dribble the ball up the court. He must control the flow of the offense. And he is responsible for distributing the ball to his teammates in places on the court where they can score. A player who can truly dominate the ball in this way may be considered a pure point guard, a point god, a floor general.
Among the bevy of stats available on NBA.com, at least three are useful surrogates for these ball-handling responsibilities: a player’s dribble count (estimated as touches per game x dribbles per touch), his time of possession (minutes per game with the ball in his hands), and his assist count (per game).
If we examine Curry’s trends in these stats since the 2013-14 season (the first with publicly-available tracking data), we see a steep decline in Curry’s ball dominance over time. This season, he’s reaching five-year lows in time of possession, dribbles, and assists per game.
To provide some context for Curry’s ball-control stats, we can match each of his last five seasons to similar performances that have been turned in recently by other players. I used hierarchical clustering (hclust in R) to group players by four per-game stats: time of possession, dribbles, assists, and minutes played. I included all seasons from 2013-14 to 2017-18 in which a player logged at least 500 minutes. I created 20 player clusters and found that eight of them contained nominal point guards with each group exhibiting varying degrees of ball dominance.
During the 2013-14 season, Curry was at the more-controlling end of the point-guard spectrum (7.4 minutes of possession, 446 dribbles, and 8.5 assists per game), finding himself in the same orange grouping as the 2014-version of Chris Paul (8.0 minutes of possession, 445 dribbles, and 10.7 assists per game).
From 2014-15 to 2016-17, Curry backslid into the blue groupings with the likes of combo guards 2014-George Hill and 2016-CJ McCollum. This season, Curry is reaching the hinterlands of ball dominance (in purple), earning himself a spot next to back-up point guards 2015-Jeremy Lin, 2016-JJ Barea, and 2016-TJ McConnell with a time of possession just a smidge above the five-minute per game threshold that tends to identify a team’s “primary ball-handler”.
Curry’s new style of play was initiated by Warriors coach Steve Kerr. Starting in the 2014-15 season, Kerr ushered in an era of increased ball and player movement for the Warriors. The strategy produced a reduction in Curry’s on-ball responsibilities and it has paid dividends for the Warriors; but it also perpetuated a perception of him as the “tip of the spear” — a player who benefits from the momentum of the system that propels him, rather than its driving force. As we’ll see, in reality, Curry is more like the arm throwing the spear than he is its tip.
Influence of Curry’s covert operations
Despite having the ball in his hands less and less, Curry’s influence on the Warriors offense has not diminished. As Ben Taylor described, here, a few months ago, the Warriors offense is supercharged with Curry on the floor (123 points per 100 possessions) and pedestrian with him on the bench (109 points per 100 possessions). That’s a bigger ON-OFF influence (+14) than any of his All-Star teammates wield, twice as much, even, as Durant.
It’s useful to consider Curry’s impact on his individual teammates as well, as Benjamin Morris did for FiveThirtyEight last year. For example, in 2016-17, nearly every Warrior shot better (based on true shooting percentage) when sharing the floor with Curry than when playing without him. We can take the weighted average of the difference in true shooting percentages for each Warrior (weighted by the harmonic mean of true shot attempts with and without Curry, using a minimum harmonic mean of 30), to summarize this effect by season. For example, last year, Curry improved his average teammate’s true shooting by 8.0 percentage points — that’s astronomical. It’s the equivalent of the difference between the Warriors’ shooting (best in the league this year at 61.0 TS%) and the Kings’ (worst in the league at 53.1 TS%).
The reason Curry’s teammates shoot better with him on the court is simple — they get easier looks when he’s around. When Curry is in the game, his average teammate is assisted on a much larger proportion of his 2-point field goals and he takes more of his field goal attempts at the rim than when Curry is out.
Moreover, when Curry is on the court, each of his teammates is expected to shoulder a smaller burden of the offensive load. This is manifest in smaller average usage rates and lower turnover rates for Curry’s teammates when he is on versus when he is off.
Curry’s beneficial influence is consistent from year to year — he has positively impacted his teammates each of the last five seasons, irrespective of coaching and personnel turnover. He’s dominating the ball a lot less this season, but he’s making the same huge impact on the Warriors offense.
Once again, we can contrast Curry’s stats with his peers’ — this time using the league’s best scorers as the comparison group (i.e., the 30 players with the highest scoring averages over the past four seasons combined). Take a look at the average improvement in teammate true shooting percentage (weighted by field goal attempts) when each star player is on the court. Curry’s impact is unparalleled. He’s responsible for both of the top two seasons, including last year’s astounding +8.0 TS%. Even his least impressive seasons (+4.7 TS%) eclipse the best efforts from all but three of the league’s other biggest stars (LeBron James, Thompson, and DeMarcus Cousins).
There is an interesting mix of positions up and down the chart. Still, most of the nominal point guards (Curry, Russell Westbrook, Paul, Isaiah Thomas, James Harden, Kyrie Irving, John Wall) have reliably positive impacts on teammate shooting, as you might expect. In contrast, wing scorers (Andrew Wiggins, Carmelo Anthony, DeMar DeRozan, Jimmy Butler, Gordon Hayward) were less consistently helpful, each having a negative impact on teammate true shooting during at least one of the last four seasons. Thompson, Durant, Bradley Beal, and Paul George offer contradictions to these positional trends; but, like Curry, each is a 3-point threat and therefore likely creates opportunities for teammates via floor spacing.
Overall, there’s not a strong relationship between a player’s own scoring efficiency (as defined by TS%) and the benefit he provides to his average teammate’s scoring efficiency. Curry is the rare star who can score super efficiently while also helping his teammates score more efficiently — he IS a combo guard, but…like, in a GOOD way.
But if Curry doesn’t have the ball in his hands, how exactly is he exerting his influence? To understand, let’s look at how he’s helping two of his teammates: Thompson and JaVale McGee.
To McGee and Thompson, Curry is a great ally
During the past two seasons, McGee has experienced a career resurgence with the Warriors — playing a small but meaningful role in the team’s success. Curry has been key to McGee’s revitalization.
With Curry on the court, McGee has attained league-leading-level true shooting — 68 percent over the past two seasons. Fully 87 percent of McGee’s shots in those situations have come at the rim. In contrast, McGee has had less success (57 TS%) while taking fewer shots at the rim (51 percent) without Curry these past two years.
Note, while Curry seems to be facilitating McGee’s scoring opportunities, he’s not necessarily ASSISTING his buckets, directly. For example, consider the most common Curry-McGee lineup — the one where the pair plays alongside starters Thompson, Durant, and Green. When McGee’s been assisted in that lineup, Curry has provided the final pass only 33 percent of the time — with the other three players also frequently assisting (Durant 24 percent, Thompson 22 percent, Green 20 percent).
This is in contrast to the classic host-parasite relationships point guards Westbrook and Harden have formed with their centers, Steven Adams and Clint Capela, respectively. Likewise, Curry’s current approach is a big shift from what he was doing with his former big man, Andrew Bogut during the 2013-14 season. Curry served up more than half of Bogut’s free meals when the pair played in the starting lineup that year (53 percent).
Unlike Adams or Capela — who are often high pick and rollers — McGee is asked to be a bystander in Green-Curry high-screen sets that Golden State runs in the halfcourt.
In these situations, opposing defenses are forced to honor Curry’s long-range threat by having Green’s defender “show out” to the 3-point line — as the Clippers’ Danilo Gallinari does here. That leaves Green with acres of space in the lane and it gives McGee a good opportunity for a dunk. Curry doesn’t get an assist for the play in this clip, but he created the 2-on-1 chance for Green and McGee by forcing the double team. Plays like these have Curry leading the NBA in secondary (or so-called “hockey”) assists this year at 1.4 per game.
This is all a bit of “small sample-size theater”, though, because McGee hasn’t played that much with the Warriors and he hasn’t taken that many shots with or without Curry. But the evidence is much more robust when considering Curry’s other beneficiary, Thompson.
Just like McGee, Thompson consistently shoots better when he shares the court with Curry. Since 2013-14, Thompson has improved his true shooting by 5, 7, 2, 7, and, now, 13(!) percentage points when paired with Curry compared to the minutes he plays without his running mate. Thompson doesn’t need to force as many tough shots with Curry around to free up space for him – as reflected by his usage rate, which dropped over the same stretch of seasons by 3, 5, 5, 5, and 7 percentage points with Curry on the floor.
But — like McGee — only a fraction of Thompson’s assisted shots are passes from Curry. Check out, for example, the assist distribution for Thompson among the Warriors starters over the past two years — just 27% came from Curry in that lineup.
At times, though, Curry’s mere presence can free up space for Thompson. In this clip, for example, Curry runs himself into Thompson’s defender, Andrew Wiggins. Curry’s defender, Jeff Teague, is in position to retreat towards the basket and help cover Thompson’s cut to the rim or, at least, to “open up” so that Wiggins can have some room to follow Thompson, himself. Instead, Teague stays glued to Curry, worried that he will pop open for a wing 3-pointer. Consequently, Wiggins is rubbed free from Thompson, leaving him to finish an easy lay-up.
In both cases — by triggering a chain reaction that led to a McGee alley-oop or by distracting defenders to open a cutting lane for Thompson — Curry gained the advantage for his teammate despite not having made an assist.
A Warrior on the mend
With Curry sidelined by a series of ankle injuries this season, we’ve caught a glimpse of what the Warriors look like without him. They’re a ho-hum 13-8 in the 21 games he has missed so far this year, including a recent slide of four losses in six games.
Now — with Curry once again unavailable for an extended period of time after spraining his knee in the Warriors win over the Hawks on Friday — we will see just how critical the point guard is to Golden State’s offensive attack.
Next: Nylon Calculus -- Joel Embiid has your sample size right here
Of course, his teammates already know how important Curry is to their success. As Durant candidly acknowledged recently, “He is the system, here.”
The Warriors are a different — much less potent — team when Curry leaves the floor. If his knee injury is slow to heal, Golden State may legitimately become an underdog to win the title. For the first time all season, the threat of a challenge to the Warriors’ supremacy is suddenly very real.