# Audit: h3k (H3000 hooked pools) - auditor, 9 Oct 2026
**Verdict: FIX / NOT VERIFIED YET.** 5 trades; look-ahead in the pool filter; arms C-F rest on 3-4 trades.

- Trades (24 rows, 4-5 per arm): A mean 3.67x, B 5.50x, C 14.1x, D 20.7x, E 27.6x, F 17.6x; 0 zeros; 1 stop/time loser per arm.
## 2. Outside sellers WITH tx.from (my run, outside.py)
A (5x): 4/4 exits, 28-145 distinct non-creator sellers, 12-111 ETH sold within +-60 s at >= 80% of target. C (20x): 3/3, 17-35 sellers, 5.6-23 ETH. D (30x): 3/3, 18-50 sellers, 6.9-24 ETH. Buyers per pool 160-1640 wallets, top wallet 0-5%, creator 0%. So 20x/30x prints are NOT one wallet: this overturns tm-audit-1.md ("C/D/F unrealistic") on the market-depth side. A2 #138 ("no outside wallet sold above 5x in 113 pools") is about the pool population, not these 3 winners: these 3 pools are survivors of a filter.
## 3. Look-ahead: FIX
24 of 43 pools are skipped as "shared hook (#131)", i.e. hook address used by a second pool. Whether a hook gets a second pool is known only after entry. Compute sibling count only from pools initialised before the entry time, and report the skipped group too (what the same rule would have earned on them).
## 4. Fills: not rebuilt; sell size 0.04 ETH vs 100+ ETH volume is small. 3% slippage + 0.3% fee as stated. Late = 0 on all rows.
## 1. Sellability: hook skim/blocks untested; real outsider sells at the target price are indirect evidence only. Needs $5 sell (Ben).
## 5. Replay vs live: none.
## Read: arms E/D (27x, 20x) are means of 3-4 pools; survivor selection (5 of 43 pools). Treat as unproven until ~30 trades.

## Re-check 10 Oct 08:40 UTC (#590, replay-h3k-pub.md, n=102 per arm)
- Look-ahead/survivorship: FIXED on the filter side. Sibling rule now uses earlier pools only; 0 pools had an earlier sibling, 0 in trades_sib. Not traded: 17 NO TAPE (no swaps at all, nothing to buy) + 19 rule filter (13 reserve<0.4). Late=0 on all 102.
- Still suspicious: arm A 98/102 hit the 5x TP (median 4.6x net, 2 pull, 2 stop/time); median time to TP 651 s, median entry reserve 2.07. A 96% hit rate to 5x is a bug-or-pump-population until shown otherwise: the pool set is H3000 pools that were selected by class, and the 17 NO TAPE pools are not shown to be dead rather than untracked. Arms E/D 33x/25x means rest on a few huge winners (zeros 16/7).
- Not done: fills hand-rebuilt from chain for 3 trades; outside sellers with tx.from for the new 102; sellability ($5 sell, #525/#528, no MATCH pool yet).
- Verdict: still NOT VERIFIED. Next: hand-rebuild 3 trades and outside-seller check when I have a slot; sellability waits for Ben's sell.