Celtics Spot-Up Three Creation Sources

Generated 2026-05-19. Answers the upstream question: what initiating actions actually created Boston's spot-up threes? Prior analyses established Spotup as the dominant ending bucket; this one looks at what feeds it.

Source endpoints retrieved this session via nba_api 1.10.2:

• LeagueDashPtStats(pt_measure_type=Drives|CatchShoot|PullUpShot|Passing|PaintTouch|ElbowTouch|PostTouch)
• TeamDashPtPass, PlayerDashPtShots, TeamDashPtShots
• Local PBP cache: data/pbp_cache_v3/0042500111-7.csv (full PHI series)

---

1. Executive verdict

Spotup is an ending bucket, not an initiating action. Public data can answer the upstream creation question only partially:

• Direct, high-confidence: receiver-side shot-type splits (catch-and-shoot vs pull-up 3PA per player) and passer-side potential / secondary assist totals per player. NBA tracking does expose these.
• Direct, partial: drive volume, drive-passes, drive-assists per player. The drive endpoint does NOT isolate "drive-pass leading to 3PA," only "drive-assist" (any FG type).
• Proxy only: the action chain. Public tracking does NOT credit the screener/passer of a specific Spotup 3PA. PBP descriptions do not label PNR / DHO / Iso, so the prior-event window is a low-confidence text proxy.
• Not measurable in public data: Queta-specific roll gravity creating kickout-threes. Best we can do is a PBP "Queta-event in prior 5 BOS events" flag (29.7% of captured PHI-series 3PA), which is suggestive only.

The data that IS available paints this picture for the 7-game PHI series: roughly half of Boston's captured 3PA followed a rebound (transition + offensive-rebound putback-3 mix), ~19% followed a drive event, ~4% followed a PNR-like sequence in description, ~8% were unassisted pull-ups, and the rest were ball-movement or cut relocations. Tatum was the largest creator by potential assists (12.2/game), Brown drove most (18.1/game), Pritchard had the highest drive-assist conversion (1.7/g on 9.6 drives).

The poster's claim ("the actions that produce 3PA are what coaching prioritizes") is partially supported: yes, Spotup is the dominant ending bucket, and Boston runs it heavily. But on the question of what creates the Spotup chance, the data shows drives and rebounds do most of the work, not pure ball-movement or pure PNR. And Boston's catch-and-shoot 3 conversion was 31.2% in the playoffs - the generation worked, the finishing didn't.

2. Why Spotup is an ending bucket, not an initiating action

Synergy classifies a possession by how it ended:

• A handler PNR that becomes a kickout-three is filed under Spotup for the shooter, NOT under PRBallHandler.
• A roll-man pass-out that becomes a corner three is Spotup, not PRRollman.
• A drive-and-kick three is Spotup, not "Drive".
• A handoff-into-three by the receiver is Spotup, not Handoff.

So when we say "Boston ran Spotup 23-25% of possessions," we're counting the chances that ended at a stationary shooter, not the actions that created them. The creation is upstream.

3. Data availability

Endpoint audit (full table in data/celtics_spotup_three_creation_endpoint_audit.csv):

Not available in public data:

• The direct chain "Spotup 3PA #N was caused by PNR action of player X with screener Y." That would require Synergy's full event-level tagging (paid) or NBA's tracking event data (not public at the required granularity).
• Roll gravity attribution. Synergy explicitly does not credit the screener for downstream Spotup 3PAs.

4. Direct tracking findings (receivers and creators)

Catch-and-Shoot 3PA per Boston player (playoffs, totals = per-game × GP)

The catch-and-shoot 3PA pool in the series was ~195+ attempts total, spread across 9 shooters. That matches the TeamDashPtShots total of 205 catch-and-shoot 3PA across the 7 games.

Pull-up 3PA per Boston player (playoffs)

Tatum took the most pull-up 3s by a wide margin. Hauser and Scheierman almost don't pull up. Brown and Pritchard are balanced.

Passing creators (playoffs)

Largest playoff creator pipeline: Tatum (12.2 potential AST/game, 18.0 assist-points created). Brown drove most (18.1/g) but converted drives to assists at the lowest rate (0.05 AST per drive). Pritchard was the most efficient drive-creator (0.18 AST per drive).

5. PBP reconstruction findings (PHI series, 185 captured 3PAs)

From data/celtics_philly_3pa_creation_context.csv. Each Boston 3PA classified by keyword scan of the prior 5 Boston events.

Creation-class summary

Read carefully:

1. 49% of captured BOS 3PA had a rebound in the prior 5 events. That bundles defensive-rebound→transition-3 with offensive-rebound putback-3 chances. This is the largest single class.
2. 18.9% followed a drive event in the prior 5 BOS events. That maps closely to the drive-and-kick narrative.
3. PNR-like text matches were rare (4.2%) because PBP description text doesn't reliably say "screen" or "pick" - it says "made 3pt jumpshot, no assist" or "(P. Pritchard 3 AST)." So PNR is undercounted by this method. The 41.7% conversion rate on the 12 PBP-detected PNR-three sequences is the best 3P% across all classes - but the n is tiny.
4. Unassisted_or_unknown (23 3PA, 17.4% 3P%) is the pull-up bucket that went cold.

Receiver PBP totals (PHI series)

PBP captured 185 of the official ~320 Boston 3PA across 7 games (the description text is incomplete; subType-only 3PAs are missed). Of the 185 captured:

Queta took 3 3PAs across the entire series. He is not a 3PA-generation source as a shooter.

6. Creation-class summary (combined direct + proxy)

The most defensible read is from triangulating three independent sources:

So the picture: drives and rebounds (transition) generated the chunkier share of Boston's spot-up looks. Pure half-court PNR creation was a smaller contributor than the volume of Spotup ending buckets suggests.

7. Player receiver summary

• Hauser and Scheierman are pure off-ball shooters. Hauser had 4.0 catch-and-shoot 3PA/g in the playoffs; Scheierman 2.1. Scheierman shot 40.0% on his CS 3PAs - the best volume-quality combination in the series.
• White was a heavy CS shooter (5.4 3PA/g, 37.8 total) at 23.7% CS FG3%. That single line is the largest individual contributor to Boston's playoff 3PA collapse.
• Tatum was a pull-up shooter, not a CS shooter (5.8 pull-up 3PA/g vs 2.7 CS 3PA/g). Most of his playoff 3PA volume came from his own PNR pull-ups, not from kickouts.
• Brown was balanced (3.0 CS, 2.9 Pull-up per game).
• Vučević and Garza are pop-roll shooters. Both took 0 pull-ups and had CS 3PA volume from popping out of PNRs.

8. Player passer summary

• Tatum was the playoff hub: 63 passes/g, 12.2 potential AST, 18.0 assist-points created. His 6 games carried most of the half-court creation load.
• Brown drove most (18.1/g) but converted at the lowest rate. Drive-to-AST = 0.05.
• Pritchard was the most efficient drive-creator (drive-AST 1.7 per game on 9.6 drives = 0.18).
• White's drive volume was low (5.1/g) for a starter. His role was spot-up reception more than creation.
• Hauser and Queta did not create. Hauser 0.9 AST/g, Queta 0.6 AST/g. Both are finishers in the passing chain.

9. Queta gravity / roll-pressure findings

This is the hardest question. What we can say:

• Direct measurement: NO. Public NBA tracking does not credit the screener for downstream Spotup 3PAs. Synergy attributes them to the shooter only.
• Lineup-based measurement: NOT available in this session. Would require leaguedashlineups filtered to Queta-on vs Queta-off combinations and the per-lineup 3PA / 3P% computation. The endpoint exists; not fetched this pass.
• PBP proxy: 55 of 185 captured Boston 3PA (29.7%) had a Queta event mentioned in the prior 5 BOS events. This is suggestive but is also confounded by Queta minutes: when he's on the floor, he appears in the recent-event window by default, even when his presence didn't cause the 3PA.
• Queta's own tracking: he passes almost nothing. 0.6 AST/g playoffs, 1.4 potential AST/g. Drives 1.3/g. Drive-AST 0.0/g. He doesn't create with the pass - he's a screener and a finisher.

Conclusion on Queta gravity: Cannot be measured in public data. The PBP "Queta-in-prior-5" rate of 29.7% is consistent with the basketball claim that his roll forced help and opened threes, but also consistent with "he's on the floor for ~25 min/game and shows up in any recent window." The data refuses both confirmation and refutation.

10. What supports the poster's claim

• "Coaches design offenses to produce 3PA." TRUE.
• "Boston runs Spotup heavily." TRUE - 27 possessions/game in the playoffs, 1st of 16.
• "Most of Boston's 3PA come from catch-and-shoot situations." TRUE - 64.5% RS, 64.1% playoffs.
• "PNR Roll Man directly generates few 3PA." TRUE - PRRollman has ~1.2 inferred 3PM/g in playoffs, almost none from Queta.

11. What does not support the claim

• "Queta on the roll is good but doesn't produce 3PA - so it's not what the staff wants." Partially false:
  • Roll gravity creates Spotup chances; Synergy doesn't credit the roller. The kickout-three IS attributed somewhere, but to the receiver's Spotup row.
  • Boston's playoff CS 3PA went in at 31.2%. The 3PA-generation strategy didn't pay off at conversion in this series.
  • Drives generated more measurable creation (Brown 18.1, Tatum 11.8, Pritchard 9.6 per game) than half-court ball-movement. Drives ARE rim pressure - which is what Queta's roll also is.
  • Brad's stated preference for first-shot offense and rim pressure maps to the drive AND roll branches, not the pure CS pipeline.
• "What the coaching staff wants is what the data shows." The data shows Mazzulla ran the CS pipeline heavily and it didn't convert. That is a critique of the choice, not a defense of it.

12. What remains unproven

• The exact share of Boston's CS 3PA that originated from a PNR pass- out vs a drive-and-kick vs a transition lead pass vs a swing-swing.
• Whether more Queta rolls would have created more CS 3PA via gravity. Plausible but not measured.
• Whether Mazzulla's stated preferences (rim pressure, first-shot offense) matched his actual play-calling. The data describes outcomes, not internal intent.

13. RealGM-ready response

The "Spotup is the action that produces 3PA" line is half right. Spotup is the ending bucket where 3PA get recorded, but Synergy doesn't tell you what created the chance. NBA tracking does help. In the PHI series, Boston took ~320 3PA and 64% were catch-and- shoot. Of the 185 we could pick up cleanly from play-by-play context, 49% followed a rebound (transition + putback mix), 19% followed a drive event, 8% were pull-ups, and only 4% had any PNR keyword in the prior five Boston events. The biggest playoff creators were Tatum (12.2 potential assists per game on 63 passes) and Pritchard (1.7 drive-assists per game on 9.6 drives - the most efficient drive-creator on the team). Brown drove most often (18.1 per game) but converted drives to assists at the lowest rate. So the actions that GENERATED Boston's 3PA were drives and transition, not pure ball-movement or pure PNR. Queta-specific roll gravity can't be isolated in public data - the screener doesn't get attribution credit - but his "involvement" (any Queta event in the prior 5 BOS events) shows up before 29.7% of captured Boston 3PA, which is consistent with the gravity claim but doesn't prove it. The honest read: Boston's offense produced 3PA mostly via rim pressure and transition, not via the ball-movement archetype the stat-sheet shape suggests. That actually aligns with Brad's stated preference for first-shot offense and rim pressure - which is exactly the branch Queta's roll lives in.

---

Sources and limits

Sources retrieved this session:

• LeagueDashPtStats for measure types: Drives, CatchShoot, PullUpShot, Passing, PaintTouch, ElbowTouch, PostTouch (× RegularSeason / Playoffs) - 14 JSONs in data/raw/tracking/
• TeamDashPtPass (BOS, RS + Playoffs) - 2 JSONs
• PlayerDashPtShots (10 Boston players × 2 season types) - 20 JSONs
• TeamDashPtShots (BOS, RS + Playoffs) - 2 JSONs (from prior fetch)
• Local PBP cache: 7 PHI series games (0042500111.csv to 0042500117.csv)

Deliverables produced:

• scripts/analyze_spotup_three_creation_sources.py
• data/celtics_spotup_three_creation_endpoint_audit.csv (39 rows)
• data/celtics_philly_3pa_creation_context.csv (185 rows)
• data/celtics_3pa_creation_summary.csv
• data/celtics_player_3pa_creation_receivers.csv
• data/celtics_player_3pa_creation_passers.csv
• data/celtics_queta_gravity_proxy.csv

Limits, declared:

• Public NBA data does NOT identify the screener of a downstream Spotup 3PA. Roll gravity is not directly measurable.
• The PBP cache description does not label PNR / DHO actions. Prior- event keyword classification is a LOW-confidence proxy.
• PBP captured 185 of ~320 official Boston 3PA in the PHI series (description-text undercount).
• Drive endpoint does not isolate "drive-to-3PA" - only "drive-to- any-FG-assist."
• Lineup-based Queta on/off analysis was not pulled this session; would require leaguedashlineups and is the next-step fix.
• 7-game playoff sample. Per-class rates are directional, not population estimates.