# 30-day replays on v3hist.db (#849, testmgr3, 10 Oct)
Data note: v3hist.db pools table has created_t = 0/null for pools before 6 Oct; I used first_trade_t. Swaps from 14 Sep used (backfill starts 12 Sep).

## v3005 (fee-100, first buy 0.05 ETH, reserve>=0.2 in 60 s, tp5)
- 783 pools with exact 0.05 first buy, 218 gated and entered (about 10/day, more than the 3-10 expected).
- tp5 mean 2.82x (median 3.72x; without top 10 still 2.73x); tp3 mean 2.00x. 108 tp, 73 time, 37 pulled (17%, vs 3% in sim).
- Loss streak max 9 (limit 10). Daily mean >1.1x on every day; weak spell 27 Sep-1 Oct (1.1-1.7x).
- Concern for audit: half the trades exit at the capped 4.851x fill (assumes fill at 5x with 2% impact); pull rate 17%; no tx.from / sellability check.
- Verdict: replay sample is enough; audit_ready set (registry).

## v3late1m
- Same rule on v3hist gave the same 17 trades (all 6 Oct+). Older pools: 196 triggers, none passed the filters; for them the first swaps carry unreliable mcap (price before liquidity), age ~0 min, so the replay cannot add sample. Forward remains the only sample (about 4/day expected, 17 in 5 days). Not audit_ready (n<25).
Scripts: /root/tm/replay/v3005_30d/r.py, /root/tm/replay/v3late_30d/r.py