# a3 iter 13 Oct (idle run, 10 iterations; scripts m1-m7.py)

## KEY: v4 native nohook fee-100 seed-25k kit (venue v4:native, cap_first 24-26k, depth ~87k at 300k, 160k at 1M): the A3-NH100 paper test is blind, and the kit converts far better than the ledger says
- Radar (live + backfill) 193 kit pools touched 300k since 30 Sep; 1M reached: 47 (24%). By day: 12-13 Oct (live feed): 22 of 45 (49%), earlier live 4 of 30 (13%), backfill 21 of 118 (18%). Not a measurement artifact (live vs backfill), not BTC/ETH (6h/24h moves flat, 87/88 split 23 vs 16 hits).
- 1M is a real price rise: depth_usd goes 87k -> 160k at the 1M cross (= sqrt(3.3) on fixed liquidity). 300k -> 1M takes 20-138 min (median ~35). Ceiling 2M (HIVE, NAP only), 0 reach 3M.
- v4tape.py header: A1 (non-hook) pools are taped only until first swap + 25 min. Kit pools touch 300k at median 23 min after first swap, so the tape ends right at the touch: of 47 kit pools that reached 1M, 0 have tape up to the 1M cross (23 NONE, 14 PRE-ONLY, 10 PART). A3-NH100 forward "quiet" exits (12 of 14 on-time) are the last print about 1.0x: the forward number (1.18x, tp 1 of 14) measures nothing. Same for its 4 h window.
- Loser leg unknown (no path). Break-even with TP at 1M (3.2x net) and a 0.45x loser leg is P(1M|300k) = 20%. Window 10 days 24% (EV ~1.1), last 2 days 49% (EV ~1.8). Regime dependent; ceiling 1M-2M so it is a flip, not a runner catcher beyond the $1M tier.
- Hot-hand inside the kit (prior 10 resolved kit pools >=30% success): 38% (23/61) vs 18% (22/123); halves not consistent (hot>=1 in first half 13% vs 21%). INFO.
- cap_5m < 80k: 9/13 reach 1M (late 7/8); opposite of #529 older sample (0/1.7/6% for <60k). Noise/regime. INFO.
- Asked ppfeeds: extend v4tape window for A1 fee-100 native/weth pools that touch cap >= 150k to touch + 3 h (all swaps). Then NH100 replay/forward is valid and a kit-300k test with real stop paths can be run.

## Other iterations
1. Block-0 kit features (kitlive/kit.jsonl, 5559 v3 pools; join radar max cross): creator buys in block 0 (b0_creator): 300k rate 2.4% vs 7.1% among born<100k fee100 (after 6 Oct: 6 of ~540 vs ~5% in no-buy), P1M 1.0% vs 1.8%; b0_recipients=0 shows 26% P1M but 59/95 are one day (7 Oct batch) = artifact. DISCARD (weak, SERIAL/SEED cover).
2. Sibling-pool confirmation (token with a 2nd pool crossing 300k within 30 min, radar): 72% vs 46% to 1M, but mirror/decoy pools of the same token inflate it (stock/usdg pinned, v4:weth siblings born at price). DISCARD (confounded, no real-vs-phantom label).
3. v2 0x8bce forward (v2trig n=18, was 15, 3 new): new sells_pre>=3 trigger 749a5c (sp 10) never reached 5x, reserve gone within 1876 blocks (~7 min). Gate sells_pre>=3 now n=3: 2 hit 5x (9af28b, f3c385), 1 pulled. Correction to run 94 note: after the 5x print the winners kept half their WETH reserve for 12.1k and 55k blocks (blocks ~4.75/s: 43 min and 3.2 h), not 100 blocks. TP is not a race for those two; the risk is the pull BEFORE 5x (entry at the 8x trigger). sp0 hit 5x once (e26028, reserve 14 WETH at 5x). INFO for A3-V2Q-S3.
4. Name/meta carry-over for the kit: name word shared with a 1M crosser of the last 24 h: 3/15 vs 44/178 (20% vs 25%). DISCARD.
5. Regime BTC/ETH: see above, flat. DISCARD.