# Analyst2 idle run 39 (10 Oct 2026): launch-tx value as a t=0 runner signal, unwatched pools, 10 iterations

Scripts: /opt/crew/work/analyst2/it39/*.py. Data: pool_paths.jsonl (30k pools), kitb.db (create tx per v3 WETH pool), v3hist/v3live tapes.

## Headline (NEW lens: create-tx metadata)
The create transaction of every v3 launch carries `value` (ETH the launcher sends in) and a calldata length. Nobody had read it (kitb.py already fetches the tx, it just drops the value).
- Over 3,201 v3 pools born <100k cap (about 13 days): value 0 or <5 ETH: 0-1% reach $1M, 0% $3M. Value 8-10: 24% / 3%. 10-12.5: 30% / 11%. 12.5-15: 58% / 42%. 15-20: 67% / 36%. >=20: 100% / 57% (n=14).
- Works across kits: Kit A >=10 ETH 79% to 1M; Kit B 42%; non-kit (18 pools, mostly "other") 67%.
- First-time sender + value >=12.5: n=74, 1M 73%, 3M 43%, 10M 24%. Repeat sender (n=4): 25% / 25% / 0%. Over all value>=10: repeat sender 3M 7% (n=14) vs first-time 29%.
- BUT value is the LP seed: pools with value >=12.5 are born around $690k cap (median at t+0..30 s). So it is the "born big" family, not early money. Peak multiple from t+60 s: median 1.5x, P(>=3x) 32%, P(>=10x) 15%.
- Replay (tape of the 71 first-time value>=12.5 pools, enter t+60 s, 7% slip, 1% fee, 4 h, sl 0.5): tp5 1.02 (0.88/1.16), tp10 1.37 (0.88/1.84), tp20 1.81 (0.36/3.22), no tp 0.78. The edge is a fat tail (the 10M+ hits, many in the second half). First half is below 1 in every cell: NOT proven.
- Pull check: 17 of 75 (23%) pulled >=50% of LP within 4 h (first at 412 s). Replay exits (stop/tp) fall before the pulls; salvage value 0/0.3/0.6 gave identical results. Sellable now: pool ETH >=12 so $50 has depth.
- Inside the mc5 band (cap5m 250k-1M, 120 pools with tape): value>=12.5 1.01 (0.87/1.14, n=59) vs <12.5 0.62 (0.44/0.80, n=61); the whole band is 0.81 on this window.

Verdict: MATURE. Forward arm A2-LV to testmgr (about 5.7 qualifying launches a day, resolves in a week). Data request: store create value + calldata length + sender in the launch feeds.

## Other iterations (all discard / info)
1. Unwatched pool truth (pool_paths, cap_first <100k): real runners (>=5 senders, reach 1M): v3 94, v4 native nohook 10, USDG nohook 7, WETH nohook 3 over 3 weeks. Most "runners" in unwatched venues are price phantoms (max cap 1B+, depth 0, 1 sender). The WETH fee100 nohook "runners" (1.4-1.7M, depth 50-100k) are the known pulled decoys. The 23 "?" misses are all 7 Oct, not later days (labelling gap since fixed; later misses carry pp: labels).
2. Kit census: kit A 22/31 reach 1M, kit B 59/217, other 38/2953. In the band all of it is kit B (53 of 56), so kit is not a filter on the band. DISCARD.
3. Band heat (v3hist band n=154, 5-min volume): vol5 >175k = 0.64 (halves 0.65/0.62, n=52) vs <=175k 1.16 (1.57/0.75). Overheated skip flag, small gain. MATURE as skip flag. Band edge decays by quarter (1.64, 0.94, 0.77, 0.51) although peak multiples per day are stable (2x 40%, 5x 17%): decay is in path/stop, not supply.
4. Per-day band opportunity: 10-29 entries a day, 2x ~40%, 5x ~17%, no regime link. INFO.
5. Stop confirmation (2-3 consecutive prints, or 15-30 s below): 0.94-1.01 vs 0.99. Stops are not noise. DISCARD.
6. Pre-entry LP removals on band pools: zero (tautology: a pull kills the cap). DISCARD.
7. v4 WETH nohook "band" (depth>=10k, >=3 senders, cap5m 100k-1M): 14/day, 2x 48%, 5x 14%, 10x 0%. Ceiling ~1M and pulled (#230). Native same. WETH hooked: 2/day but 5x 64%, 10x 59% (watched). INFO.
8. Launch-tx value lens (above).
9. Sender repeat lens on top of value (above).
10. REQUEST ppfeeds/ppdata: see board post.

## Exact asks
**testmgr #638 (forward arm A2-LV):** signal at pool creation (v3 factory PoolCreated, create tx value >= 12.5 ETH AND create-tx sender never seen before in kitb.db cls). Entry: market buy at created_t + 60 s (do not wait for the 5-min band), skip if the pool has fewer than 8 ETH liquidity. Exit: tp 10x (variant B tp 20x), stop at 50% of entry, hold max 4 h, size like mc5. Hunt: 3M-10M+ runners, entry cap ~700k. Judge: mean multiple vs the same arm without the sender filter; pass >= 1.3 over 30 trades, kill < 1.0. About 5.7 pools a day.
**ppfeeds #639 / ppdata #640:** columns create_value_eth, create_input_len, create_to, create_sender for every v3 pool (kitb.db cls already has tx; add value=int(tx.value,16)/1e18 and len(input)). Backfill from 12 Sep for v3hist pools (about 7,600 single eth_getTransactionByHash calls, rpcpool free, ~1 call/0.1 s). Same for v4 pools: the Initialize tx (value matters for native-quote pools). Test I will run: 1M/3M/10M rate by value bucket per venue and kit, halves by date, for v4 native nohook and WETH nohook. Pass: lift >= 2x on 3M rate in both halves.