Celtics Prosperity Analysis, All Years (2021-22 through 2025-26)

Generated 2026-05-11. Multi-year extension of celtics_tatum_brown_prosperity_analysis.md and celtics_tatum_brown_baseline_context_addendum.md.

Coverage: 81 Celtics playoff games across 14 playoff series, 5 postseasons (2021-22 Finals run, 2022-23 ECF, 2023-24 championship, 2024-25 R2 loss, 2025-26 R1 loss). 11,394 tagged event rows. Built from nba_api.stats.endpoints.playbyplayv3.

Sources:

• data/celtics_all_years_tagged_events.csv (raw events)
• data/celtics_all_years_context_summary_by_player.csv
• data/celtics_all_years_context_summary_groups.csv
• data/celtics_all_years_delta_vs_baseline.csv

---

Definitions (unchanged from prior reports)

• Prosperity possession (lead_8plus): period >= 3, Boston margin >= 8 entering the event.

• Smaller-lead (lead_5plus): period >= 3, margin >= 5 entering event.

• Non-prosperity pressure: period >= 3, margin between -5 and 0 inclusive.

• Second-half-all: period >= 3, any margin.

• Full-series: every BOS tagged event.

• Ender: the BOS player who finished the possession (FGA, FT, TO).

• Creator: the BOS player credited with the assist on a made BOS shot.

• all_roles: ender + creator rows. Double-counts assisted Boston makes.

• ender_only: just enders. Cleaner for possession-value comparison.

• Force event (proxy): rim attempt OR free-throw attempt OR assisted three created. Public-data approximation; not Second Spectrum.

---

Sample sizes (ender_only, Celtics team total)

Tatum prosperity sample by season: 131, 102, 150, 59, 27 (469 total). Brown prosperity sample by season: 88, 107, 117, 55, 28 (395 total).

---

Table 1. Tatum, prosperity vs full-series baseline by season (ender_only)

Read: Tatum's prosperity turnover rate is up vs his own baseline in 4 of 5 seasons. The 2025-26 +7.19 pp surge is the dominant case. The underlying pattern is consistent (TO up in 4 of 5) but the magnitude outside 2025-26 is modest (+1 to +2.4 pp). Force rate drops in 4 of 5 seasons. Rim rate drops in 4 of 5 seasons. PPP drops in 2 of 5 (the ECF year and the championship year).

Table 2. Brown, prosperity vs full-series baseline by season (ender_only)

Read: Brown's prosperity midrange rate is negative in 3 of 5 seasons vs his own baseline. The only seasons with positive midrange-in-prosperity deltas are 2023-24 (+0.82, small) and 2025-26 (+8.61, large). His turnover rate goes down in 4 of 5 seasons in prosperity. His prosperity PPP is roughly flat at multi-year level. The 2025-26 "Brown settles in prosperity" finding does not generalize.

Table 3. Tatum vs Brown vs Other Celtics, all-years combined (ender_only, full series baseline)

Three structural findings jump out:

1. The Jays are less efficient than their teammates in every context. Across 5 seasons of playoff data, Other Celtics produce 0.964 PPP over 5,046 ender events. Tatum and Brown combined produce 0.875 PPP over 4,439 ender events. The gap is ~0.09 PPP and is consistent at the season level.

2. The gap widens in prosperity. Other Celtics improve to 0.990 PPP in prosperity. The Jays combined drop to 0.859 PPP. Gap widens to ~0.13 PPP.

3. Force rate is the opposite story. The Jays have higher force rate (~44-46%) than Other Celtics (~34-38%) because they take more rim shots and draw more fouls. Other Celtics are catch-and-shoot threats (40-44% three rate) who produce fewer force events per possession but more efficient possessions.

Table 4. Tatum_Brown_combined vs Other Celtics, by season (ender_only, prosperity vs full series)

Read: The Jays-vs-Other-Celtics gap in prosperity widens markedly in 2022-23 (ECF year, Tatum's worst Jays-led prosperity efficiency), 2023-24 (championship year), and 2024-25 (NYK loss). It closes in 2025-26 because both Jays were efficient in their small prosperity samples, while Other Celtics were also efficient.

Table 5. Strongest evidence series

Series where the Tatum-TO-surge pattern is strongest:

Both ECF losses and the Stevens series show the same pattern. Tatum playing the ECF series against MIA in 2022-23 (Boston blew a 3-2 ECF lead) showed a +6.55 pp TO rate spike in prosperity. Combined with the 2025-26 +7.19, the two most prominent "let them back in" series in the dataset have the highest Tatum TO surges.

Series where Brown's prosperity-PPP is weakest:

Brown's worst prosperity series at meaningful sample size is the 2024-25 second-round NYK loss (32 events, 0.375 PPP). That is a major signal that Brown specifically broke down with leads in a series Boston lost. The 2025-26 series, despite the famous narrative, was a positive prosperity series for Brown (1.071 PPP).

Table 6. Series that weaken the simple "Jays in prosperity" thesis

The 2024-25 ORL first-round series is the strongest single counterexample: both Jays improved in prosperity, both reduced turnovers, and both scored above their full-series PPP. The prior 2022-23 PHI series (the one preceding the 2025-26 PHI rematch) also has Tatum reducing turnovers in prosperity, which directly weakens "Tatum is the prosperity-TO problem."

Table 7. Sample-size table by season and context (ender_only)

Single-series prosperity samples are small (8 to 57 events). Multi-year samples are robust (469 Tatum, 395 Brown, 1034 Other Celtics in prosperity).

---

Q1. Does the Tatum prosperity-specific turnover surge replicate across years?

Yes in direction, no in magnitude. Tatum's prosperity TO rate is above his full-series baseline in 4 of 5 seasons (2021-22 +1.02 pp, 2022-23 +2.38, 2023-24 +1.06, 2025-26 +7.19; 2024-25 was -1.73). The multi-year delta is +1.13 pp on 469 prosperity ender events. He is the only one of the five groups (Tatum, Brown, Jays-combined, Others, Team) whose TO rate increases in prosperity vs full-series; everyone else decreases. So the directional finding replicates.

What does not replicate is the +7.19 pp magnitude from 2025-26. That single-series spike is roughly 6x the multi-year average. Treat 2025-26 as the extreme tail of an underlying pattern, not as the typical case.

The 2022-23 MIA ECF series (also a "let them back in" collapse from a 3-2 lead) shows the second-largest Tatum prosperity TO surge at +6.55 pp, which is the closest analog to 2025-26 in the dataset.

Q2. Does Brown's midrange-heavy shot diet replicate across years?

The diet itself replicates. The prosperity-intensifies pattern does not.

Brown's full-series midrange rate is 20.08% across 2,136 ender events, roughly double Other Celtics' 11.87% and Tatum's 16.02%. That is a real durable pattern: Brown takes more midrange shots than his teammates in every season and every context.

But the specific claim that prosperity makes Brown settle more is not supported across years. Brown's prosperity midrange rate is lower than his full-series midrange rate in 3 of 5 seasons (2021-22 -1.19, 2022-23 -1.31, 2024-25 -5.64). The two seasons where it goes up are 2023-24 (+0.82, small) and 2025-26 (+8.61). At multi-year level Brown's midrange-in-prosperity is actually -1.09 pp vs baseline.

The 2025-26 "Brown settled more with leads" finding is a one-series artifact.

Q3. Are Tatum and Brown both worse in prosperity?

Both have lower force rate and lower rim rate in prosperity. PPP is roughly flat for the pair.

Multi-year deltas (ender_only, prosperity vs full series):

• Tatum: PPP -0.03, rim rate -2.62, force rate -2.40.
• Brown: PPP -0.00, rim rate -1.88, force rate -0.52.
• Combined: PPP -0.02, rim rate -2.34, force rate -1.53.

Net: yes, both go a little softer in prosperity, but the PPP cost is small and largely absorbed by Tatum's TO surge specifically. Calling both "worse in prosperity" overstates a small effect for Brown.

Q4. Are they worse than other Celtics teammates in prosperity?

Yes, clearly. ~0.13 PPP gap on a 1,000+ event sample.

In prosperity (ender_only, all years):

• Other Celtics PPP: 0.990 (1034 events).
• Jays combined PPP: 0.859 (864 events).
• Gap: 0.131 PPP.

This is the strongest, most replicable finding in the entire dataset. Other Celtics are more efficient in prosperity than the Jays in 4 of 5 seasons (the exception is 2025-26 where both groups were efficient on small samples).

That said: the gap is partly structural. Role players are catch-and- shoot. Stars are shot creators taking the hardest shots. Comparing PPP directly across roles overstates the "Jays problem" because it does not account for shot difficulty.

A cleaner read of the same numbers: when Boston has a lead, role-player threes are the efficient way out and Boston shifts toward that diet team-wide (3PT rate climbs from 33.44% baseline to 36.25% in prosperity). The Jays' share of those threes is much smaller than the role players'.

Q5. Is this a Jays problem, a team-wide problem, or a one-series artifact?

Mostly team-wide, with one Tatum-specific signal that survives across years.

Team-wide patterns visible in 4-5 of 5 seasons:

• Rim rate drops in prosperity (-2.27 pp team-wide).
• Force rate drops in prosperity (-2.78 pp team-wide).
• Three-point rate rises in prosperity (+2.81 pp team-wide).
• Turnover rate drops in prosperity for everyone except Tatum.

Stevens' "lacks rim pressure" and "slows down" complaints are visible team-wide in the rim-rate and force-rate decreases. Those are not Jays-specific.

Tatum-specific patterns visible in 4-5 of 5 seasons:

• Turnover rate increases in prosperity. He is the only player group whose TO rate goes up when Boston gets the lead. The magnitude is small most years (+1 to +2.4 pp) but the 2025-26 series (+7.19) and 2022-23 MIA ECF (+6.55) are extreme cases.

Brown-specific patterns: none survive across years. His high midrange rate is a constant, not a prosperity effect. His prosperity TO rate goes down in 4 of 5 years.

One-series artifacts (claims that depend on 2025-26 alone):

• "Brown settles more in prosperity than at baseline." Specific to 2025-26.
• "Brown's prosperity PPP collapses." 2024-25 NYK was the worst case (0.375 on 32 events); 2025-26 was actually a positive prosperity series for Brown.

Q6. Which seasons/series are the strongest evidence?

For the Tatum-specific TO-in-prosperity thesis:

• 2025-26_BOS_PHI (Stevens series): +7.19 pp Tatum TO delta, 27 events. Smallest sample but largest effect.
• 2022-23_BOS_MIA (ECF collapse from 3-2 up): +6.55 pp Tatum TO delta, 26 events.

For the team-wide softening thesis:

• 2023-24_BOS_MIA (R1 series in the championship year): T+B PPP -0.16, Other PPP -0.05, both groups went soft.
• 2022-23_BOS_ATL: T+B PPP -0.13, Tatum rim rate -5.79.

For the "Jays worse than Other Celtics" thesis (the most robust finding):

• 2022-23 (all series): gap = 0.205 PPP (Other 1.014, T+B 0.809).
• 2024-25 (NYK series specifically): gap > 0.30.
• 2023-24: gap = 0.178 PPP.

Q7. Which seasons/series weaken the claim?

• 2024-25_BOS_ORL: both Jays improved in prosperity. Tatum TO rate fell to 0% on 27 events; Brown PPP rose to 1.130. This series alone cuts against every star-specific prosperity-failure claim.
• 2022-23_BOS_PHI: Tatum's TO rate dropped (-3.59 pp) in prosperity. He played the previous BOS-PHI series clean. This weakens any "Tatum has a Philly problem" narrative.
• 2021-22 Finals run: both Jays' prosperity PPP improved vs baseline, TO rates dropped, midrange dropped. The team lost the Finals, but the prosperity stretches were not where the loss happened.

Q8. What can public data prove?

• Per-event classification: rim / three / midrange / FT / turnover with high accuracy (0% unknown classification across 11,394 events).
• Score margin entering each event (within the limit of forward-filled scoreHome/scoreAway).
• Which player finished the possession and which player got the assist.
• Within-player rate changes across context buckets, on multi-thousand- event samples (Tatum 2,303 events, Brown 2,136 events, others 5,046).
• Robust between-group comparisons (Jays vs Other Celtics) at multi-year level.

Q9. What can public data not prove?

• Possession initiator. A Brown FGA could be Tatum-created and the data doesn't say.
• Shot quality (defender distance, openness, contestedness).
• Shot-clock state at the moment of the shot.
• Whether a turnover was forced or unforced.
• Lineup-level effects.
• Whether team play-calling shifts in prosperity (the rim-rate drop and three-rate rise could be either player decisions or coaching).
• Free-throw trips vs free-throw events. This dataset counts each FT as an event, so a two-shot trip counts twice.
• Counterfactuals: what would the same series have looked like with a different ball-handler? Public PBP can't answer.

Q10. Cleanest honest message-board conclusion

The simple "Tatum is the common thread" framing is partially supported on the dimension Brad Stevens specifically cited (giving the opponent life). Across five postseasons of public play-by-play, Tatum is the only player whose turnover rate climbs when Boston gets the lead. The other four player groups (Brown, the rest of the roster, the team total, and the Jays as a pair) all turn it over less in prosperity than at baseline. That is a Tatum-specific signal that survives 469 prosperity events.

The "Brown is the pre-existing settle-for-midrange problem" framing from the 2025-26 analysis does not generalize. Brown's midrange diet is constant across contexts. It does not intensify with leads in any season except 2025-26.

The largest, most consistent finding in the dataset is not Tatum-vs-Brown at all. It is that the rest of the Celtics outscore the Jays in prosperity by ~0.13 PPP per possession, in every year except 2025-26. That is partly structural (role-player threes are efficient), partly real (the Jays carry the hardest shots), and partly the thing Stevens was actually describing when he said Boston "lacks rim pressure": the rim pressure that does come is from the Jays, and it falls off slightly in prosperity windows.

Honest verdict on the original thesis options:

1. Tatum thesis supported: partially. Turnover-specific only, with 2025-26 as outlier.
2. Brown thesis supported: no. 2025-26 was a one-series artifact.
3. Both are the problem: no, not symmetrically.
4. Jays structure is the problem: partially. Stars do harder shots and convert less than role players in lead-protection windows, but that's a structural feature of the offense, not a failure mode.
5. Team-wide issue: partially supported. Rim rate and force rate drop team-wide in prosperity. The "Boston goes soft with leads" pattern is real but it shows up in pressure proxies, not in PPP.
6. Inconclusive: partly true on small-sample series, false on multi-year aggregates.

If forced into one sentence: the prosperity collapse Stevens described is team-wide softening of rim pressure, with Tatum as the marginal turnover contributor in 4 of 5 years and Brown contributing nothing distinctly different from his normal play.

Caveats

• Public PBP does not identify true possession initiator.
• FT events are attempts, not trips. A two-shot trip counts as two events.
• Shot clock not available.
• Defender distance not available.
• all_roles view double-counts assisted Boston makes (one row for ender, one for creator). ender_only view is cleaner for possession value.
• Single-series samples are noisy (8 to 57 prosperity ender events per series).
• 2025-26 series sample (130 team-wide ender events in prosperity) is smaller than every other season; its high PPP (1.038) may be a small- sample artifact.
• "Force event" is a public-data proxy, not Second Spectrum drive count or quantified shot quality.

Bottom line for a 2-line message-board take

Across five postseasons of public play-by-play, Tatum is the only player group whose turnover rate goes up when Boston has the lead; everyone else (Brown, the rest of the roster, the team) turns it over less. That signal is small on average (+1 pp) but extreme in two specific "let them back in" series (2025-26 PHI +7.19 pp, 2022-23 MIA ECF +6.55 pp). Brown does not show a prosperity-specific failure mode at the multi-year level. The bigger structural fact is that the rest of the Celtics outscore the Jays by ~0.13 PPP in prosperity windows on 1,000+ events. That gap is Boston's actual prosperity problem, and it predates 2025-26.