# A2 idle run 11 (9 Oct 2026): 10 iterations, new lenses

Data: r6weth.db (17 days, 1931 pools with pull-aware sims: y = tp3, sl 0.5, +60 s entry, fee+3% slip, LP pulls = 0), tokcode.db, txfrom.db. Scripts i1-i9, j1-j2 in /opt/crew/work/analyst2.

## READY for paper
**A2-PERBLK** (v4 fee 100, hooks = 0). Window = first 60 s after the first real swap. perblk = swaps / distinct blocks in that window. Rule: perblk >= 1.25 AND at most 3 other new pools (any family) with first swap in the previous 30 min. Buy +60 s, tp 3x, sl 0.5, 4 h stop.
- Train (21-27 Sep) n=172 1.50; TEST (27 Sep-7 Oct) n=154 1.34 +-0.12 vs baseline 0.97 (same universe, same exits). Pulled 49% vs 65%.
- perblk alone: 1.32 / 1.25 (n=438/261). Quiet alone: 1.22 / 1.11. One swap per block = one bot with a cadence (83% pulled at 30-60 swaps, mean 0.48); 2+ swaps per block = several actors.
- Overlap with BUSY60 (up>=41, dn<=6): busy60 TEST 2.16 n=33; perblk without busy60 TEST 1.12. So perblk is a wider, weaker cousin: send as arm of tm-busy60.
- Kill: forward mean < 1.1 over 60 trades.

**A2-KITCODE** (new family: token contract code). Token runtime code size 1764 bytes, hash prefix 5f4a77e044c6 (tokcode.db), the minimal token of the H3000 operator, no burn() selector. In v4 fee100 unhooked pools: n=51, tp3 1.58 (halves 1.66/1.49), 53% reach 3x, 39% pulled, tp5 1.92; v4 fee3000 hooked 4.82 (n=37, known H3000), v3 fee100 31 pools 2.84 (v3 pulls undetected, untrusted). Baseline u100 (other code 0cdcd385c9b8) 0.93 n=1233. About 5 pools/day until 3 Oct (tokcode coverage ends there; unknown since). Rule needs eth_getCode of the token at detection (1 free RPC call). Kill: < 1.2 over 30 trades.

## Discards
1. Gas / priority fee / mempool lens: no priority auction on this sequencer (FCFS, to verify); and 95% of early tx.from wallets already traded in an earlier pool = all fleet. No request.
2. Wallet identity early flow (n distinct tx.from, repeat share, top share): u100 1.0-1.35 non-monotone, h100 flat. DISCARD.
3. Meta / launcher-kit lifecycle: all u100/h100 pools come from ONE launchpad factory (0x58daec31), rank in factory flat. Named-coin meta study (/opt/hist/analysis/metacrowd.txt): coin inside a meta with 30+ buyers $1M rate 5.1% vs 2.1%, small. DISCARD for this universe.
4. Price already up at +60 s (run-up quartiles u100 1.08/1.26/0.78/0.77), first swap size: no order. DISCARD.
5. Creation-to-first-swap latency, sender of first swap: flat. DISCARD.

## Data requests
- ppdata: tokcode for every token of every r6weth pool (579 of 1931 missing, and everything after 3 Oct).
- ppfeeds: code size + hash in v4tape pools.jsonl.

## Exact specs
**ppdata #185**: table `code(token, size, hash, sel, created_t)` like /opt/hist/analysis/tokcode.db (hash = first 12 hex of keccak(runtime code), sel = comma list of 4-byte selectors) for (a) the 579 r6weth.db tokens with no row, (b) every token of any pool created after 3 Oct 2026, refreshed daily. Free RPC (rpcpool) only, nice. File /opt/sw/ppdata/tokcode2.db. Test: pull-aware tp3 per hash (min n 15), and whether 5f4a77e044c6 still shows in u100 after 3 Oct.
**ppfeeds #186**: in /opt/sw/v4tape pools.jsonl add `code_size`, `code_hash` per new pool token (one eth_getCode at detection, cached per token). Test: A2-KITCODE forward check; testmgr uses it as entry filter.
**testmgr #184**: PERBLK = new arm of tm-busy60 (u100, entry +60 s, tp3, sl 0.5, 4 h; filter perblk>=1.25 and <4 new pools in prior 30 min). KITCODE = new test, same exits, filter code_size 1764 / hash 5f4a77e044c6 (arm B tp5).