Every good series transforms at some point from a mystery -- the teams exploring lineups, feeling out schemes -- into something firmer. After that threshold, both teams know the fundamentals: who they want to play and when; how they want to attack on offense; their defensive principles against those attacks.
The series becomes about execution, toughness, poise, and a little luck.
These strange, riveting NBA Finals aren't quite there, but they're approaching the strategic endgame. Rotations have been whittled, schemes tweaked and tweaked again. One team has made a starting lineup change.
The aggregate score is 422-421, Warriors. Both teams have 64 3s on solid-to-great shooting. The Golden State Warriors have one more offensive rebound than the Boston Celtics and one fewer turnover. The Celtics are plus-7 in free throws.
The bookend games -- Games 1 and 4 -- were similar: road teams roaring to wins behind rampaging fourth-quarter runs. The middle games were relative blowouts.
The league's No. 1 and 2 defenses have won the day; both teams are scoring about 110.5 points per 100 possessions, equivalent to the 20th-ranked offense in the regular season.
As expected, Boston feels steadier, with enough reliable support to withstand Jayson Tatum's miserable 2-point shooting; he's 14-of-51 on 2s, and struggled again in Game 4 to strike the balance between decisiveness -- as both passer and scorer -- and flailing into traffic. Golden State's offense outside the Stephen Curry maelstrom feels precarious, though Andrew Wiggins has done his job and the Warriors found ways to spring Klay Thompson in Boston.
Curry has 137 points; Golden State's No. 2 scorer, Thompson, has 69. Curry is averaging 34 points on 50% shooting -- including 49% from deep -- against a defense that obliterated everyone once Boston turned its season in January. That defense -- and Golden State's inconsistent scoring around Curry -- is stretching him to his limits.
Curry's pick-and-rolls have increased every game, per Second Spectrum tracking: 30, 35, 39, 45. That Game 4 number was his second highest for the season. Curry has attempted only 11 catch-and-shoot 3s -- and a preposterous 40 pull-ups. (That split was closer to even in the regular season.) He's averaging only 3.8 assists. Boston's half-court defense has shut off the Golden State beautiful game -- the whir of cuts and splits and back screens in which Draymond Green lives -- that usually produces oodles of easy assisted buckets.
(Curry is 20-of-40 on those pull-up 3s, which is straight-up absurd. Robert Williams III and especially Al Horford were up higher corralling Curry in Game 4. They moved their feet, and navigated two and three screens in rapid succession. Curry's defenders -- Marcus Smart and Derrick White -- hounded him from behind. Freeze some of those pull-up 3s, and Curry is rising in a window of space barely wide enough to contain his body. He's in a phone booth. What a remarkable series for him so far.)
Williams' mobility might end up a decisive factor. Boston is plus-20 with Williams on the floor. He handled the Green assignment fine when Golden State started Green at center in Game 4. (I had expected Williams to guard Otto Porter Jr., with Horford on Wiggins and Tatum on Green. But coach Ime Udoka trusted Williams, and it worked.)
Boston has found good balance between going small -- with four guards and wings around either Horford or Williams -- and leveraging its size with two-big lineups. The right mix varies game to game, but Udoka has a good feel for it.
Four Boston lineups have appeared in all four games. Remarkably, only one Golden State group has done so: their former-and-maybe-present-again starting five, with both Kevon Looney and the frantic Green, a group that is plus-22 in 50 minutes. Looney is a team-best plus-36. Golden State has rebounded a gargantuan 37.5% of its own misses with Looney on the floor, scrounging non-Curry points. That group finished Game 4, and Steve Kerr might want to start it in Game 5 (Monday, 9 p.m. ET, ABC) -- hoping Green found his rhythm during a productive final few minutes in Game 4.
Regardless of which lineup starts, I wonder if Thompson's work guarding Jaylen Brown in crunch time might nudge Kerr toward shifting Green back onto Horford or Williams.
Green has done solid work on Brown, but the effect of his snarling intensity seems to have waned. Green being around the rim, in help position, is generally good for the Warriors' defense. This was one reason the Gary Payton II-Wiggins-Green trio appeared promising in Game 2; Payton and Wiggins defended Tatum and Brown, sending Green back onto a Boston big.
The lineup of Curry, Payton, Wiggins, Porter, and Green helped swing Game 2. It logged two combined minutes in Games 3 and 4. The Payton-Green combination made for clunky spacing, and both Brown and Tatum are growing more aggressive shooting over Payton.
Kerr is plugging holes, reaching for any semblance of two-way balance. In the fourth quarter of Game 4, with Golden State's season teetering, Kerr turned to a three-guard lineup of Curry, Jordan Poole, Thompson, Wiggins, and Looney -- benching Green, and banking on a group with 78 minutes of prior shared time. That is really a modified version of the once-vaunted Poole Party small-ball group, with Looney in Green's place; the actual Poole Party has logged eight minutes in four games. (It will be fascinating to monitor Green's minutes going forward.)
Kerr even toyed with super-duper-small-ball, briefly playing Curry, Poole, Payton, and Thompson together. It seems late now to toss Jonathan Kuminga or Moses Moody into the fray, but perhaps Kerr might dare it in a home game. There will be adjustments we don't see coming.
The Celtics let Poole off the hook down the stretch of Game 4 as their offense unraveled again. It wasn't the turnovers this time, though Boston committed 16 -- 10 of which were steals. (Forty of Boston's 60 turnovers have been of the live-ball variety that ignites Golden State's fast-break fireball. Only 27 of Golden State's 59 cough-ups have been live-ball steals.)
All game, Boston appeared rushed and addled for stretches. The Celtics drove headlong against Golden State's best defenders when one simple action would have produced something more favorable. They jacked pull-up isolation 3s early in the clock -- Smart (22% on pull-up 3s this season) over Looney, Payton Pritchard going hero ball in semi-transition, Brown missing almost everything after dancing with Nemanja Bjelica.
The Thompson trey that gave Golden State the lead for good came after Brown rejected a pick from White -- with Poole on White -- to go at Thompson one-on-one. Thompson stuck to Brown, and forced an ugly floater. Use the pick, and make Poole work. Tatum missed a contested layup earlier doing this same thing -- with Wiggins reading Tatum's intent to reject the pick, and staying on his hip.
Compounding things, Boston's spacing went awry on some of those drives. Watch White and Williams cut into the paint as Tatum revs up to get there:
Even a couple of Boston drive-and-kick 3s in crunch time -- one miss from Tatum, another by Horford -- were pretty well contested because the drives that produced them came against Golden State's top defenders, and didn't generate enough traction to really suck in the help. (Don't overlook fatigue -- mental and physical -- as one possible driver of these puzzling late-game stretches from Boston. The Celtics' top players have logged a ton of taxing postseason minutes now.)
When the Celtics honed in on their game plan, they mostly got what they wanted. They have leaned on the Tatum-Smart two-man game to generate mismatches for Tatum all postseason. They redirected some of those pick-and-rolls into off-ball actions with the same effect. Tatum and Brown would amble into some off-ball action with whichever Celtic had Curry or Poole guarding him: pin-downs, flare screens, even Tatum screening underneath the rim for Pritchard.
Golden State resisted switching those actions even though it has largely been OK switching Curry and even Poole onto Tatum in the pick-and-roll. Any non-switch defense -- chasing over picks, shooting the gap, whatever -- opens gaps, and the Celtics found good looks in those gaps.
The Warriors' willingness to switch Curry onto Tatum and Brown in pick-and-rolls has been one of the more surprising subplots, though I'm not sure it surprised Boston's coaches. It appears Golden State would rather give Boston a size mismatch than put two on the ball -- as it did when Luka Doncic dragged Curry into pick-and-rolls -- and ignite Boston's passing game.
The Warriors want to force Boston to play one-on-one, in the mud. They are betting on their ability to flash help, confuse Boston's stars, and force the turnovers that power their fast breaks. (That said, the Warriors could be more diligent going under picks when Smart is the ball handler and Tatum is the screener.)
The Celtics in Game 3 grew comfortable attacking those switches. When Horford slid to center, they slotted everyone around the arc -- opening the lane for Tatum's and Brown's spinning, bulldozing drives through Curry and Poole. When Williams was in the game, he hung around the dunker spot during those drives -- a lob threat to occupy the last line of defense. Tatum and Brown digested Golden State's reluctance to help off Williams, and kept plowing to the rim.
When help came early -- as it generally has at the nail off White, Horford, Smart and Grant Williams -- Boston's stars dished early. The Celtics have hit 44% on catch-and-shoot 3s. Their role players have shot well enough for them to take this thing.
The Celtics lost the thread on all this for much of Game 4. They haven't really figured out what to do against Looney on the perimeter either; there are methods to go at him. Boston needs to play with focus for 48 minutes to beat this Warriors team twice more.
The other big question is whether Boston should adjust its pick-and-roll defense against Curry -- to have the defender on Curry's screener venture up even farther, or to outright trap him. I've seen strong arguments both ways: Of course you blitz! Did you See Game 4? And: Don't overreact, the Warriors are scoring at a below-average rate and they've been bad in the half court; Boston's issues are on offense.
The question is more nuanced than all that, and may require more nuanced answers. The Warriors are scoring 116.9 points per 100 possessions with Curry on the floor -- almost two points above the Utah Jazz's league-leading figure. Their half-court scoring numbers are indeed poor, but a lot of their pick-and-rolls for Curry come near midcourt in semi-transition -- when Boston is backpedaling.
(This is also true of some plays in which Thompson rockets off pin-downs from whoever is being guarded by Boston's center:
Boston does not want to switch those actions, so they always open up something.)
Still: I'm not sure a dramatic adjustment on Curry would be smart, nor that we'll see one.
The Warriors scored only 0.846 points per possession when Curry came off a ball screen and did something that ended Golden State's possession -- i.e., shot, turned it over, or passed to a teammate who launched. That was their lowest single-game figure of the series. Include entire trips featuring a Curry pick-and-roll, and the Warriors scored only one point per possession -- average-ish.
That seems counterintuitive given the fireworks, but Curry committed five turnovers in Game 4 -- several out of pick-and-rolls. When we hear "drop coverage," we tend to think of someone like Brook Lopez hanging in the paint. In Game 4, Boston's drop coverage looked like this:
Horford is 2 feet above the arc. If Curry's defender slithers around the pick unscathed, that's high enough to deter anything but ridiculous-degree-of-difficulty 3s. I mean, look at this defense -- White's pursuit, Horford's positioning, the way White and Brown trade assignments on the fly:
(Smart is on Green there, and one benefit of Boston going smaller is shifting Smart to that assignment. He has been fantastic helping off Green to clog the lane, and then flying back outside if Green gets the ball and pivots into one of his dribble handoffs.)
When Curry has tried to drive by Horford, Boston's rotations have been mostly on point:
Curry can handle blitzes blindfolded if he knows they're coming. The all-out assault on Curry risks bringing Green back into the series:
Blitzing with both Green and Looney on the floor is one thing. Golden State's spacing is tight enough that fast, smart defenses can get back into those plays as Green and Looney ping the ball around. It gets tougher -- the passing lanes wider, the reads easier -- with four shooters around Green.
I suspect the Celtics' defense in Game 5 will look broadly similar, perhaps with a few blitzes and even switches on Curry mixed in. Switching is another area where Williams' mobility is paramount. Curry has had a harder time solving Williams' combination of quickness and extreme vertical shot-blocking. He hasn't had much trouble roasting Horford, and switching unlocks those classic Curry give-and-go relocation 3s -- one of which came in Game 4 when Horford switched onto Curry but wasn't sure whose job it was to continue the chase:
Golden State could also dial up the complexity, running staggered screens, rescreens, and its patented full-speed screen-the-screener play -- in which a third Warrior slams Green's guy with a pick while Green runs to screen for Curry:
There is a ridiculous amount of tactical stuff and full-throttle execution going on in that seven-second clip. The level of play in this series is sky high -- the flexibility and smarts of these two defenses, and the creativity they require of opposing offenses.
That's what the Finals is supposed to be. Boston is the better, deeper team on paper, but the Warriors reclaimed home-court by seizing upon the precise sort of inconsistency that makes the Celtics a hair more vulnerable than they should be.
The series is now something close to 50-50 toss-up. Game 5 will demand even more of both these great teams.