# KILL: the 10-second flip (A2-FLIP, A2-FLIPFEAT, A2-FLIPGATE, A2-FLIP180). FW Analyst 2, 8 Oct 2026, 16:30 Bangkok

## Verdict
Outside money gets rugged. The paper test cannot see this, so it should be stopped (Ben decides). Do not go live.

## Evidence (launchpad 0x58daec3116aae6d93017baaea7749052e8a04fa7, fee 1%, 9,452 coins, Sep 20 to Oct 7)
- Closed loop: 235 wallets that each trade 20+ of these coins do 99.81% of all buy ETH. Outside wallets (in fewer than 20 coins) put in 0.04%. The "real trades" the backtest was priced from are the operator's own fleet.
- After an outside buy, the coin is pulled within 5 s: 81% (under 0.002 ETH), 88% (0.002-0.005), 67% (0.005-0.02), 34% (0.02-0.05), 22% (0.05-0.2). Baseline for a random second: 0.78%. Fleet buys of any size: about 1%.
- 2,607 of 3,526 outside buys came less than 5 s before the pull.
- Coins alive at 15 s with an outside buy of 0.003-0.05 ETH at 5-15 s (n 13): 69% pulled by 30 s. Without one (n 8,296): 5%.
- Every outside position (n 3,326): nothing back 89%, mean 0.12x. Our size ($10, about 0.004 ETH) bought in the first 20 s: nothing back 93%, mean 0.10x, profitable 3%.
- The pulls go through the operator contract 0x1cbaf24d53fe930fce8eff149fa797d2611da149 (used in 285 of 289 of these coins vs 140 of 2,711 other pools; it also routes the fleet's buys and fans tokens out to about 100 wallets at launch).

## Why the earlier checks missed it
- Real-trade pricing used the fleet's trades, which never trigger the pull.
- The earlier "outside wallet" label came from crew.db (deployer and funder wallets of other crews), i.e. other operators, not retail.

## Knock-on
- L1 (pond crew skim) and any book that buys plain v4 launches in the first minutes: run this same outside-buyer test before paper. Scripts: research-box /root/fwa2/edge/outside3.py (events with the real wallet), outside5.py (pull after outside buys), outside6.py (what outsiders got back).
- Live test 2 (AI picker, real lane) trades v4 pools too: run the same test on its coins before it goes live.