# analyst3: bundled launches vs runners on the full tape population (#437, #440), 10 Oct 2026
Data: ppdata seen_peakcap.csv (src=e1x, 5,238 pools, 17 Sep-9 Oct, clean sample) joined to bundle_fund.db bundle_launch (159 bundled pools, 138 creators, 7 big crews by funding cluster). Scripts: p1.py, p2.py (this dir). Bundle is detected in the LAUNCH block, so the flag is causal at entry.

## 1. Population result (replaces my sample-based 2.2x lift in a3-bundle-gmak-10oct.md)
Share of pools reaching peak cap: all 5,238 / unbundled 5,079 / bundled 159
- >=300k: 12.9% / 10.8% / 81%
- >=1M: 5.7% / 4.5% / 41.5%   (9x)
- >=5M: 2.3% / 1.9% / 17.6%   (9x)
- >=10M: 1.7% / 1.4% / 11.3%; >=20M: 1.3% / 1.1% / 6.3%
Bundled = 3.0% of pools but 22% of all 1M+ runners (66 of 297) and 23% of 5M+ (28 of 123).
Conservative check (thin-tape pools dropped, cap_flag=ok only; thin bias flatters the unbundled floor): 1M 43% vs 14% (3x), 5M 18% vs 3.9% (4.7x). Real lift is 3-9x, not the 2.2x of the oversampled set.
Many pools "never reach" 100k only because tape is thin; treat absolute rates as approximate, the ratio as the signal.

## 2. Dose: bundle USD size matters (n=53 each tercile)
- <$28k median $18.5k: 1M 15%, 5M 3.8% (no better than base)
- mid $28k: 1M 40%, 5M 9%
- top tercile (>~$34k, median $40.6k): 1M 70%, 5M 40%
Wallet count showed no dose effect; dollars do. Also creator NOT in the bundle (n=39): 1M 56%, 5M 28% vs creator in bundle (n=120): 37% / 14%.

## 3. Crew / funding-chain flag (#440 part b)
7 crews (>=7 launches) all behave as runner-makers, not traps: 1M rate 14-75%, 5M 7-31% (c_5cf061d6 n=16: 69% / 31%; c_5159d0ef n=12: 75% / 25%; c_e1f227ad n=49, biggest, weakest: 24% / 12%). Causal check on prior outcomes within a crew: after at least 1 earlier 1M hit the next launch hits 1M 40% (n=134); first 3 launches of a crew 59% (n=37). So crew history adds nothing beyond the bundle flag: no "good crew" list needed, and no crew is a skip candidate on peak data alone.
Caution: this is PEAK cap. Rug crews pump then dump (bundle sells median ~2 min, see earlier report), so a ride entered late loses even when peak is large. Needs entry-multiple replay: the flag says "will run", not "buyable at profit".

## 4. Pons V2 tax-exemption (#440 part a): cannot test
No exemption config/events in pons.db (only launch/trade/grad on the curve). The 159 bundles are Uniswap-pool block-0 multi-buys, not Pons tax data. Ask dataeng for Pons V2 hook "tax exempt" setter events if worth it; the Wazz/Bitquery crews' signature (70-200 wallets, 70% supply) is already inside our bundle flag (up to ~46 wallets/crew-root).

## Signals handed on
- A2 (late $300-500k hunt): bundled launch AND bundle > ~$34k is a runner-rich pool (70% reach 1M, 40% reach 5M). As an entry universe: ~1% of pools. Test as a watch-list/filter; measure entry multiple from $300-400k (needs per-pool price path from e1x tape).
- A1 (early hunt): bundle known at launch block, usable at second 0.
- Contradicts A2's old proxy (>=4 front buys = AVOID, 1h horizon): peak is reached over hours; a replay of exit timing is the open question.
## Next
Replay: bundled+large entries at 300-400k vs matched unbundled, multiple achieved with a 2-min bundle-sell exit vs hold. Needs testmgr/analyst2 tape replay.