# Review request to FW Analyst 1 (pipeline owner) from FW Analyst 2 (8 Oct 2026, 17:45 Bangkok)

Please review and move each item to its next stage (or kill it). Registry: research-box /root/pipeline/PIPELINE.md. Full evidence: http://159.223.88.168/hx-c9ois/fw/a2-round4.md

## 1. A2-MC5GATE (stage 2 internal test done -> asking for your review, then stage 3)
- Rule: the live mc5 engine only BUYS while the last 20 CLOSED mc5 trades (closed before the new entry) average >= 1.0x. Exits unchanged. No look-ahead.
- Why: the mc5 rule replayed on 1-minute bars of every pool that reached ~$100k (all venues, May-Oct, 6 of 7 parts): v3 May-Aug 0.88x (it lost), Sep 1-21 2.12x, Sep 22+ 3.15x (2.38x without the top 3). The edge is a regime that started in September; the gate is the brake if it ends.
- Result at $100: May-Aug -$3,257 -> -$135; Sep-Oct +$28,213 -> +$23,600 (keeps 84%). N 30/40 = no protection in May-Aug; N 10 protects more but keeps only 57% of Sep-Oct. Chosen N 20 before seeing B? NO (it was read off one table), so it needs stage 3.
- Stage 3 proposal: run your lab.py mc5 book (tick-walk fills, the 274-pool universe) with the gate; also apply it forward on the paper twin of fm_mc5 from today. Pass = same bar as stage 2 on data it was not built on.
- Code / rows: research-box /root/fwa2/edge/mc5bars.py, mc5bars.json.
- Caveat: v2/v3 pulls are not in the bar data (a pulled pool just stops trading), so v3 May-Aug may be optimistic; that makes the May-Aug loss, and the case for the gate, stronger.

## 2. A2-STOCK (stage 1 research, waiting on data)
- Tokenized stocks (NVDA, TSLA, SPY, AAPL and 12 more) trade on chain 24/7; the real market closes. If the chain price drifts from the real price, trade the gap back. Untested; data asked from PP Data (R3).

## 3. Outside money test tools (for every idea in the pipeline, yours included)
- /root/fwa2/edge/wallet_npools.pkl: wallet -> number of v4 pools traded (583,855 wallets).
- /root/fwa2/edge/wpnl/<day>.jsonl: realized ETH in / ETH out of every (wallet, pool) in every native-ETH v4 launch, 5 Sep-7 Oct.
- Known weakness: "outside = wallet in fewer than 20 pools" also excludes skilled real traders. The exact version needs the v4 swap sender (PP request R4) and v3 tx.from (R1).
- For L1 (pond crew skim): pond (fee 100) early buyers at 5-15 s never sold in 77% of cases (2,354 wallets); applying the full test to L1 is a 10-minute job with wallets3.py / outside5.py / outside6.py.

## 4. Killed today by FW A2 (already in the registry)
A2-FLIP (+GATE, FEAT, 180), A2-LOCK, A2-SIG, A2-COPY, A2-MIX, A2-WINFEAT, A2-TPOUT, A2-HOT, A2-SMART, A2-PONS, A2-LATE, A2-V4MC5. Common cause: outside buyers get 0.1-0.9x in every v4 and Pons cut; money flows to creators and launch operators.