# wlink: factory 0x73991a25 + deployer txcount<=5 replay (research, 16 Oct, #2681 / #2683)
Data: 41 wlink BUY rows (live + watch, wl_wlink and wl_wlink1 in fm_mirror, since 12 Oct), pool creation tx from v3live.db, deployer nonce at the block before creation (eth_getTransactionCount). Script rug5/g.py.
- Factory: 41/41 bought pools were created via 0x73991a25 (it is the shared launcher contract, tx.to). It cannot separate anything.
- Deployer nonce at creation: 29 have nonce 2, 6 have nonce 3-5, 6 unreadable (RPC). Nobody above 5. Every deployer is a one-shot fresh wallet, as in r60 and the 3 rugs of 16 Oct.
- Gate "factory & nonce<=5" blocks 35 of 41 (8 of 10 winners >=2x). Blocked mean 1.20x, kept (n=6) 1.86x. It is the whole strategy, not a filter. nonce<=2 blocks 29 (6 winners, mean 1.10x), kept n=12 mean 1.76x: small n, no edge proof.
- Zero (pulled) rate: nonce 2 = 20/29 (69%), nonce 3-5 = 4/6. No gradient.
Verdict: gate not usable, do not set rug_gate on it. #2683 pattern: the 3 rugs are 3 different one-shot nonce-2 deployers via the same launcher, same L fingerprint (already refuted as gate #2626), shared signal wallets. Only blockable thing is signal wallets (exec already added 4 to bad_wallets). Deployer wallet block is useless: next rug is a new wallet.
Suggestion: the exposure lever is size/stop rule (3-pull stop already trips), not a pre-buy gate. Funder history (who funds the nonce-2 wallets) untested; blockscout is CF-blocked, funder asked ppdata.