# A1 #749/#764: v2 swaps + tokmeta test of V2DEC9 (analyst1, 10 Oct)
Data: /opt/ppcopy/v2swaps.db (copy of 12:18 UTC; tx_from filled for the 64 dec9 + 199 ctl pairs, only 19k of 109k 'all' rows). Script v2s1.py (reserve replay from swap nets, buy 0.1 ETH at +8 s, sell at +353 s or WETH<50% max, 0.3% fee, pull=0). Pitfall: token/WETH side differs per pair (WETH side = the one that grows).

## (a) Metadata (2,426 seed-1 pairs with tokmeta + replay)
- Rule explained by DECIMALS/SUPPLY only: decimals 9 (supply 1e9, n=71) mean 39x, median 26x; decimals 18 (n=2,355) mean 2.86x, median 2.90x.
- Owner renounced: no effect (dec9 renounced 42.8x vs not 34.1x; others 3.7 vs 3.2). Code hash: one hash = 80% of pairs, no split.
- Name keywords (inu, pepe, dog, cat, frog, moon, ai...): no effect (dog/cat/frog are ~400 pairs each = kit templates; means 2.7-3.0 vs 2.9). The name is not the driver; the 9-decimals kit is.
- Replay reproduces the earlier sim: dec9 n=64 mean 38.5x at 0.1 ETH (earlier 22x, entry slightly earlier here), ctl 2.7x (= known V2RAMP arm).
## (b) Outside money (dec9 n=66 with tx_from; fleet = wallet in >=3 pairs, 1,316 of 6,178 wallets)
- First 60 s: median 0 outside buyers, 0 outside ETH; 51/66 pools have <3 outside buyers. Those 51 return 43.9x mean (median 31) vs 25.8x for pools WITH outside buyers. So my pre-set kill ("<3 outside buyers -> edge gone") is the wrong way round: the edge is highest where the fleet is alone. Early buy count also correlates up (28.6 / 42.1 / 50.7x by tertile).
- Reading: it is a fleet-driven pump (1 ETH pool, fleet buys 10+ ETH in). We ride the fleet; no outside-demand story. Control pools have 5 outside buyers and 2.7x.
## (c) Exit
- After +350 s sellers are 81% (median 86%) wallets that bought before: the fleet unwinds itself. 350 s exit is before this. Pulls ~0-3% (2 of 64 below 0.5x; 1 zero).
## Stability
- Halves 24.8x / 55.2x; 17 days with trades, ~4/day, 10 of 17 days mean >35x; worst days 3.7-12x. Forward paper test (tm-v2dec9) 5 of 5 profitable, mean 47x.
## Verdict
Keep the dec9 paper test; metadata path gives nothing beyond reserve=1e18 (decimals 9). Risk is the fleet changing its script, not outside demand. Not ready for live: needs the real-sell eth_call check, and sizing <=0.1 ETH (impact). Remaining: tx_from for 'all' group only matters for the name/outside tests, which are answered by dec9/ctl; no new data request.