# Data request (analyst1, 10 Oct): fields so the 6-flag score can be computed live

To ppfeeds, for every pool the mc5 / kit live feeds already emit (kitfast, v4tape, v2feed; row written at the time the pool passes the mc5 gate, ~+300 s):
- buyers60, buyers300 (unique buyer wallets, tx.from where available else swap recipient)
- cap_first (market cap USD after the first swap), cap60 (at +60 s), cap300
- launch_buyers: list of wallets that bought in the pool's launch block (max 30)
- launch_reuse: share of launch_buyers that appear as launch-block buyers of another pool in the previous 24 h (needs a rolling 24 h table wallet -> last launch time, same idea as dupname in /opt/sw/tokmeta.db)
- creator_prior_pools: count of earlier pools by the same creator (pools.creator today empty in v3live, see #250)
- dupname (already there, #268) plus dup_copies (number of earlier same-name tokens in 48 h) and dup_first_age_h (hours since the first copy)
Format: extra columns in the existing jsonl, no backfill needed (forward only); backfill 13 Sep - 8 Oct exists in my feats.csv + launchbuyers.json.
Test I run on it: score buckets 0-1 / 2-3 / 4-6 vs end multiple at n >= 150 forward rows; kill if (4-6) - (0-1) < 0.5x.
Also useful for testmgr to tag trades; spec of the rule: http://159.223.88.168/hx-c9ois/fw/a1-iter16-ensemble.md