Three hundred regular-season wins. Four consecutive trips to the finals. Back-to-back championships. The Golden State Warriors have enjoyed unprecedented success over the last four-and-a-half seasons. And, yet, even the best team in the league has room for improvement.
Coach Steve Kerr wants to make sure that the Warriors maintain their edge on the competition, so he’s continuing to experiment. One wrinkle this season has been the changes he’s made to his substitution patterns. Kerr has been playing around with new ways to deploy his two biggest weapons, Stephen Curry and Kevin Durant.
“The whole idea is to stagger Steph and KD the best we can. We didn’t do it much the last couple years, but we had a different team. This year’s team, we feel strongly we have to stagger.” Kerr told reporters prior to a game against the Houston Rockets on Jan. 3.
In this chart, I’m defining the Warriors “Stagger Percent” for a given game as the fraction of minutes that Curry played WITHOUT Durant on the court or vice versa (averaging between the two scenarios). During a stretch in December and January, the Warriors played twelve straight games with a Stagger Percent above 20 (in the orange box). However, for much of the season, Curry and Durant have abided by the Buddy System (in the green boxes); playing less than 20 percent of their minutes separately.
Here’s an example of the typical Curry-Durant substitution pattern that the Warriors have used during the past few years. In this game — a one-point win over the Utah Jazz — Curry and Durant played 33 minutes together, Curry played 5 minutes without Durant, and Durant played 7 minutes without Curry. So, on average, each one played 6 of 39 minutes without his co-star, equating to a 15 percent stagger.
In some ways, the Warriors situation is unique. No other NBA team has a pair of league MVPs – like Curry and Durant — who could play together (or separately). But, on the other hand, making lineup decisions like these is a universal conundrum. Every team must decide if they want to stagger stars or play them at the same time.
We can survey how the league handles staggering; but, first, we need to identify the offensive engines who are powering each team’s attack. We’ll use Ben Taylor’s load statistic to find the two players from each team who have been involved with the most scoring opportunities this season (via field goal attempts, free throw attempts, assists, turnovers, or extra spacing).
I sorted the teams by how much they stagger and then lumped them all into four groups. Of the combos, Curry and Durant have had the 10th-lowest Stagger Percent this year, at 18 percent. They are joined in the Buddy System group by Jrue Holiday and Anthony Davis of the Pelicans, Reggie Jackson and Blake Griffin of the Pistons, and Andrew Wiggins and Karl Anthony-Towns of the Timberwolves. Each of these pairings tends to spend most of their time together, but there are a handful of other couples that are even more inseparable.
Among the duos with Stagger Percents of 15 or less — those most-closely tethered binary stars of the NBA universe — we find the Wizards’ John Wall and Bradley Beal, the Blazers’ Damian Lillard and CJ McCollum, and the Thunder’s Russell Westbrook and Paul George.
Here’s a typical Westbrook-George substitution pattern from a four-point win over the Hornets. Westbrook and George played 35 minutes together, Westbrook played just 2 minutes without George, and George played 4 minutes without Westbrook. So, on average, each one played 3 of 38 minutes without his co-star, an 8 percent stagger.
On the other end of the spectrum, are the teams who like to stagger stars so that at least one of their offensive engines can be on the court, at all times. Pairings like Ben Simmons and Joel Embiid in Philadelphia as well as Eric Bledsoe and Giannis Antetokounmpo in Milwaukee have had Stagger Percents between 21 and 50 percent this year, as they frequently anchor separate units.
But the quintessential modern stagger-ers are the Houston Rockets. During the past two seasons, excluding garbage time — i.e., any time there was a margin of 15 points or more in the 4th quarter — the Rockets have played just 12 minutes with both James Harden and Chris Paul on the bench (in games when both were playing). Here’s a typical substitution pattern:
There’s another group of tandems with even larger Stagger Percents than Harden-Paul, each at 50 percent or more. These are the instances in which a team has a player like Lou Williams or J.J. Barea; a super-sub who doesn’t start but dominates the ball once he gets into the game. There’s also a handful of struggling teams — the Knicks, Hawks, and Bulls — who’ve had one of their offensive pillars in and out of the starting lineup, which confuses the Stagger Percent stats.
So, how does Kerr — or any other NBA coach, for that matter — decide? Should he stagger his stars or continue with the buddy system? Should he lean into the double-MVP lineups or, perhaps, move KD to the bench?
Well, we can look at net ratings for various lineup combinations — with both stars sharing the court together, with the stars playing separately, and with both stars on the bench — to figure out how to optimize rotation patterns. Again, we’re going to exclude data from garbage time as well as any games that either player missed. Finally, to pump up the sample size, we’re combining multiple seasons of data.
Over the past two years, Westbrook and George have been a powerful pair, outscoring opponents by 8 points for every 48 minutes they share on the court. Yet, when they’ve played separately, they end up giving back some of that advantage. And, because the Thunder lineups with neither Westbrook nor George have been at least as good as the lineups with one of the two; Oklahoma City’s best bet is to maximize the amount of time they spend together.
In contrast, Houston’s Harden-led units without Paul, as well as their Paul-led units without Harden, have been really effective. In fact, those lineups have actually outperformed lineups that feature Harden and Paul playing together over the past two seasons. As such, it makes perfect sense that the Rockets stagger Harden and Paul and avoid having them both off the court at the same time.
For some teams, the two most creative players aren’t complementary. In Brooklyn, D’Angelo Russell and Spencer Dinwiddie are capable point guards who have both demonstrated that they can lead the team, but they haven’t fit well together in the same backcourt. They drop about a point for every minute they share the court, so having Dinwiddie fill the role of super-sub is definitely the way to go for the Nets.
As for the Warriors, the optimal staggering solution is less obvious. In the past three seasons, lineups with Curry and Durant have been the bee’s knees, outscoring opponents by 18 points per 48 minutes. But lineups with either Curry or Durant haven’t been that bad either.
Assuming both Curry and Durant will play 35 minutes per game, we can use two extreme test cases to make some simple rotation calculations: (Option 1) 35 minutes with both Curry and Durant plus 13 minutes with neither (0 percent stagger) vs. (Option 2) 22 minutes with both Curry and Durant, 13 minutes with Curry only, and 13 minutes with Durant only (37 percent stagger). Based on the net ratings shown in the chart above, if Kerr were to take Option 2 (staggering Curry and Durant), it would yield an average margin of victory of 11 points. Impressive, but not optimal. With Option 1 (keeping Curry and Durant synched), Kerr could expect an average margin of victory of 13 points!
As if the Warriors lineup options weren’t already unfair enough; now – with the recent addition of Boogie Cousins – the Warriors will have yet another focal point around which to center the second-unit offense. Kerr has hinted that having Cousins to share the offensive load will allow Curry and Durant to play together even more than they have to date. That’s good news for the Warriors and bad news for the rest of the league.