Defensive Bleed After Boston Prosperity Events

Question: When Boston goes into lead-protection mode and softens its offense, does that behavior degrade the defense on the very next possession?

Window: All five Celtics playoff runs, 2021-22 through 2025-26. 81 playoff games discovered, 56 contain at least one prosperity event.

Prosperity event definition: A Boston FGA, free throw, or turnover event in the second half or later (period >= 3) with Boston ahead by 8+ entering the event.

Sample: 1,898 prosperity ender events. 1,878 (98.9%) matched to an opponent next-possession outcome within the same period.

---

TL;DR (Lead-with-uncomfortable read)

Team-wide, there is no defensive bleed. Opponents score 0.966 PPP on the possession immediately after a Boston prosperity event. That sits well below the typical playoff offensive PPP baseline (~1.10), so the defense actually performs above water in these spots.

The bleed exists, but it is player-specific and turnover-specific, not team-wide:

1. Tatum prosperity events leak defense. 1.028 opp PPP next possession. 12 percent higher than Other Celtics' 0.941. Tatum turnovers in particular cost 1.21 PPP next trip.
2. Star turnovers are the only true defensive bleed signal. Tatum and Brown turnovers in prosperity both bleed 1.21 opp PPP next trip. Role-player turnovers (Other Celtics) bleed only 0.81 PPP, well below average.
3. The "long rebound from missed three" theory is rejected. Missed threes produce 0.988 opp PPP, almost identical to the team baseline. Rim and midrange misses bleed more.
4. Missed rim attempts are the worst non-turnover bleed source at 1.083 opp PPP, ahead of missed midrange (1.013) and missed three (0.988).
5. Stevens' "softening" narrative is partially right. It is true that Boston's force rate drops in prosperity and that turnovers are the costliest event for defense. But the broader claim that lead-protection offense lets opponents back in is not supported team-wide. The mechanism is specifically Tatum's prosperity turnovers and missed jumpers.

Verdict: Turnovers create defensive bleed, but missed threes do not. The bleed is concentrated in Tatum prosperity events, not team-wide. This is verdict #3 in the analytical menu, with a Tatum-specific qualifier.

---

A. Does Boston's prosperity shot profile create defensive bleed?

Not team-wide. Boston gives up 0.966 PPP on the possession after a prosperity event. League playoff offensive PPP runs in the 1.10 to 1.14 range, so opponents underperform their own season-long output immediately after a Boston prosperity event.

The narrower question of which Boston event types are worst is where the actionable answer lives. See table 1.

B. What Boston event types lead to the highest opponent next-possession PPP?

Table 1. Opponent next-possession PPP by Boston event type (team-wide)

The order is missed rim > turnover > missed midrange > missed three > made shot > free throw event. The "lead-protection long-rebound" mechanism that the question implies is not the dominant story.

C. Are turnovers materially worse than missed threes?

Slightly. Turnover: 1.022 PPP, 90 events. Missed three: 0.988 PPP, 423 events. The gap is 0.034 PPP per possession, a small absolute difference, but the turnover rate IS elevated above the team baseline of 0.966.

The headline finding is not that turnovers in aggregate are catastrophic. It is that star turnovers are catastrophic. See section E.

D. Are missed threes worse than rim attempts or free throws?

No. Missed three (0.988 PPP) is slightly worse than free throw event (0.882) but actually better than missed rim (1.083). The cleanest opp-PPP reset comes after Boston scores or attempts a free throw, where the inbound is dead-ball and Boston gets to set. Missed threes give opponents a deeper rebound, but the data does not show that translating into more next-possession scoring.

E. Does Tatum's prosperity turnover problem create measurable defensive damage?

Yes. This is the strongest single finding in the analysis.

Table 2. Opponent PPP after Tatum prosperity events (by Boston event bucket)

Tatum turnovers in prosperity cost 1.21 opp PPP on the next trip. That is 0.19 PPP above the team baseline, on a 28-event sample. The opponent score rate after a Tatum prosperity turnover is 53.6 percent, eight points above the team baseline. That is real defensive damage.

A second flag: after Tatum makes a rim attempt in prosperity, opp PPP is 1.222. That looks counterintuitive (made shots normally produce dead-ball resets and good defense), so it deserves a note. The dataset is 54 events. Possible mechanism: a made Tatum rim attempt in lead-protection mode is often a "settle" possession where the offense slows down further and the defense reset is unrushed but also less aggressive. Plausible but not provable from public PBP.

F. Does Brown's midrange / missed-jumper profile create measurable defensive damage?

Less than Tatum's. Brown's overall prosperity bleed (0.957 opp PPP, 392 events) is below the team baseline. The exceptions:

Table 3. Opponent PPP after Brown prosperity events (by Boston event bucket)

Brown's missed midrange is actually good for the defense (0.897 PPP). Brown's missed three is the bleed source for him (1.149, 67 events), not the midrange. This contradicts the implicit framing of the original question, which expected Brown's midrange game to be the defensive liability.

The Brown-turnover signal (1.211 opp PPP, 19 events) is consistent with the Tatum-turnover signal: when a star coughs the ball up in lead protection, the defense pays.

G. Is the defensive bleed team-wide or player-specific?

Player-specific.

Table 4. Opponent PPP after Boston prosperity event, by player group

The turnover-bleed differential between stars and role players is the cleanest split in the dataset. Tatum and Brown turnovers in prosperity bleed 1.21 PPP. Role-player turnovers bleed 0.81 PPP, below the team baseline. The most plausible (but public-PBP-unprovable) mechanism: star turnovers happen in dribble-attack or pick-and-roll situations that become live-ball transition for the opponent. Role-player turnovers are more often offensive fouls, illegal screens, or dribble loss in less aggressive setups.

H. Does this support the theory that Boston's offense impacts its defense when protecting leads?

Partially. Three specific claims survive the data:

1. Boston's star turnovers in prosperity damage the defense. Tatum and Brown turnovers cost ~1.21 opp PPP next trip, on combined 47 events. This is consistent with Stevens' "letting opponents back in" framing.
2. Boston's force rate drop in prosperity is real and well-documented, but the data here shows non-force events are NOT meaningfully worse for the defense (0.964 vs 0.961 PPP for force events). The defensive cost is not driven by Boston stopping forcing the issue.
3. The lead-protection three-point reliance is defensively neutral. Missed threes produce 0.988 opp PPP, statistically indistinguishable from the team baseline of 0.966.

The broader narrative that lead-protection offense systematically damages the defense is not supported at the team level. It is supported as a Tatum-specific finding.

I. Season-by-season

Table 5. Defensive bleed by season

Key season-level reads:

• 2021-22 prosperity defense was excellent (0.916 team PPP). Boston's prosperity defense was its best in the championship-shy first-round run.
• 2023-24 (championship year) team prosperity defense was average (0.973), but Tatum's prosperity bleed was the worst of the five-year window at 1.148 opp PPP. Boston won the title anyway. This is consistent with the broader point: the bleed is real but not catastrophic at the team level.
• 2024-25 is the only season where team prosperity defense crossed 1.000 PPP (1.030). Notably, Tatum was actually the cleanest player that year (0.847); the bleed came from Other Celtics (1.053) and Brown (1.148).
• Turnover-bleed signal is consistent across seasons, but small-sample years (2022-23, 2025-26) produce extreme values like 2.25 to 2.40 PPP after turnovers, which is sample noise rather than a stable estimate.

Required: Small-sample warning table

Table 6. Sample-size warnings

Findings tagged "directional only" should be read as hypothesis generators, not conclusions. The robust headline numbers are the team-level event-type splits in Table 1 (90 to 624 events per cell) and the all-seasons Tatum/Brown/Other comparison in Table 4 (392 to 1,021 events per cell).

I. What can public data prove?

• Boston gives up 0.97 opp PPP on the next possession after a prosperity event. That is below the typical playoff offensive baseline.
• Boston's WORST defensive event type is the missed rim (1.08), followed by turnover (1.02) and missed midrange (1.01). The BEST is the free throw event (0.88) and the made shot (0.95).
• Tatum prosperity events bleed defense at 1.028 PPP, ~9 percent worse than the team baseline. Tatum prosperity turnovers bleed at 1.214 PPP, ~26 percent worse.
• The bleed is concentrated in star turnovers: Tatum and Brown prosperity turnovers bleed identically (1.21 PPP), while Other Celtics prosperity turnovers DO NOT bleed (0.81 PPP).
• The lead-protection-three-pointer theory is not supported. Missed threes produce 0.988 opp PPP, indistinguishable from the team baseline.

J. What public data can NOT prove

• Live-ball vs dead-ball turnovers. Public PBP marks an event as "Turnover" but does not flag whether it was live (live for the opponent to run) or dead (whistle, inbound, set defense). The 1.21 vs 0.81 PPP star-vs-role-player turnover differential is suggestive of a live-ball cause but cannot be directly tested.
• True transition vs half-court. Public PBP does not include a transition flag. Time-elapsed-since-prior-event is a weak proxy and was not used in the headline numbers.
• Defender distance, shot clock state, rebound length, defensive pressure. None of these are in public PBP. The "long rebound from missed three" mechanism cannot be measured directly; the rejection of that theory here is based only on next-possession scoring outcomes, which is a downstream proxy.
• Set defense vs running back. Public PBP does not identify whether Boston got matched up before the opponent crossed half-court. The "defensive reset" proxy in this dataset is purely a function of the Boston event type (made shot = inbound = likely set; turnover = likely run), not a direct measurement.
• Free throws are events, not trips. Some BOS "free throw events" are mid-trip (1 of 2, 2 of 3 etc.), not trip endings. The script handles this on the opponent side (only the last FT of a trip terminates the possession) but on the BOS side the FT event is treated as one possession ender, which is the same convention used in the broader prosperity dataset.
• Multi-event sequences (technical fouls, double-personals, jump balls) sometimes break the "next opponent possession in the same period" rule. 20 of 1,898 prosperity events (1.1%) had no matched opponent next possession within the period and were dropped from PPP calculations.
• Single-season point estimates for narrow buckets (e.g., Tatum 2022-23 turnovers, n=5) are below the threshold for meaningful inference and should be read as flags, not facts.

---

Methodology

For each Boston prosperity ender event (role = ender, period >= 3, in_prosperity_lead_8plus = True, event_type in FGA/FT/Turnover), the script walks forward in the PlayByPlayV3 stream from that event's actionNumber + 1 looking for the first opponent action event (Made Shot, Missed Shot, Free Throw, Turnover). The opponent possession is defined as ending at the first opponent FGA, opponent Turnover, or opponent's last FT of trip in the same period. Opponent points are accumulated from the BOS event's successor through the opponent possession-ending event.

This is a public-data proxy for opponent next-possession PPP. It does not perfectly handle offensive rebounds (which can extend Boston possessions past the prosperity event), technical foul interjections, or end-of-period boundaries. 1,878 of 1,898 prosperity events matched (98.9%). The 20 unmatched are end-of-period prosperity events with no opponent possession completing in the same period.

Reference baseline: NBA playoff team offensive PPP runs in the 1.10 to 1.14 range. The team-total opp PPP of 0.966 in this dataset compares favorably to that benchmark and is the basis for the "no team-wide bleed" headline read.

---

Files

• data/celtics_defensive_bleed_after_prosperity_events.csv — event-level rows, 1,898.
• data/celtics_defensive_bleed_summary_by_event_type.csv — Boston event bucket x defensive reset proxy.
• data/celtics_defensive_bleed_summary_by_player_group.csv — Tatum / Brown / Other / Team total, with per-bucket splits.
• data/celtics_defensive_bleed_summary_by_season.csv — Season x player group, with force-event and turnover splits.
• data/celtics_defensive_bleed_run.log — Append-only run log.
• scripts/build_celtics_defensive_bleed_after_prosperity.py — Reproducible build.
• data/pbp_cache_v3/<game_id>.csv — PlayByPlayV3 cache.

Generated 2026-05-11.