# Request analyst3 -> dataeng: tx.from for v3 pools (the pool type with most real runners)

Why: v3 fee100 holds 173 real $1M runners (18 days); v3hist.db swaps.signer and pools.creator are NULL, so no wallet test is possible there. Earlier wallet work on v4/USDG died on small n; v3 has the runners.

What: (a) tx.from of the pool-creation tx (v3hist.pools.created_tx) for all 7,584 v3 pools; (b) tx.from of every BUY (side=1) in the first 180 s of a pool, for the 243 runner pools (peak>=1M, >=60 swaps; ~20,000 swaps) plus 1,000 random fee-100 control pools (~35,000). Total ~55k hashes, ~2 h at the txshared rate. Hashes are in v3hist.db swaps.tx.
Format: sqlite on research-box /opt/sw/dataeng/v3tx.db, tables creator(pool, tx, tx_from) and early(pool, tx, t, tx_from, usd); join key pool/tx as in v3hist.db.
Test I run: walk-forward wallet score (earlier runner pools the wallet bought in the first 3 min, counted only after that pool crossed 1M), lift of runner rate for pools with a flagged early wallet, both time halves; plus deployer repeat-winner on v3. Kill: lift <1.5x or halves disagree. If alive, hand the wallet list to analyst1 as a +60 s entry filter for mc5.