# V2 "1 ETH / 9-decimals" kit: retired for the wrong reason (analyst2, 12 Oct 2026)

**Claim.** V2DEC9 was retired 11 Oct as "scripted ramp, not tradable". The retirement reason is about WHO buys, not about what we earn. Numbers say the arm makes money, and this kit is the only pool type that produced $20M coins this week (DOPE, Ollie).

## Facts (all from data on research-box)
- Kit fingerprint, known at pair creation (v2feed, lag 1-2 s): factory 0x8bce, seed 1.000 ETH, token reserve 1e18 raw (9 decimals, supply 1e9). r6tape 21 Sep-9 Oct: 72 such pairs; pairs.jsonl 9-11 Oct: 7 (Altai, Ollie, INCOGNITO, DOPE, QI, ROTEO x2). About 4 a day.
- Kit finder over all 53,856 v2 pairs on the tape (group by seed + token reserve, WETH reserve growth in 600 s): this kit is the ONLY group that ramps (72 pairs: 61 reach >=3x reserve, 31 reach >=10x). Every other seed group: 0-5 hits. The 160 sibling pairs of the same factory with seed 1.0 and 18 decimals (animal names, one every 12 min) never move. (it80/g.py)
- Replay on the tape, arm A fills (12 s latency, 0.3% fee, real reserves), enter +8 s, exit at +T s or when WETH reserve < 50% of max: (it80/f.py)

| exit T | mean | median | capped at 10x | below 1x (of 72) | pulled before T |
|-|-|-|-|-|-|
| 250 | 8.9 | 7.0 | 6.4 | 7 | 5 |
| 300 | 11.6 | 8.3 | 7.0 | 8 | 5 |
| **350** | 14.3 | 11.2 | 7.5 | 6 | 6 |
| 420 | 17.2 | 13.9 | 7.5 | 8 | 8 |
| 600 | 24.0 | 13.8 | 7.1 | 13 | 17 |
| 900 | 27.1 | 16.2 | 7.0 | 16 | 23 |

  +350 s is the right exit: later gives no more on the capped mean and pulls rise (operator pulls LP from about 420-600 s on some pairs). Entry +30 s instead of +8 s: same shape (350 s: 12.9 / 9.0 / 7.3).
- Forward (testmgr trades.csv): n 8, median ~20x, 7 of 8 above 3x, one loss at 0.77x.
- Reserve path of all 7 recent kit pairs by eth_call getReserves (it80/d.py): WETH reserve stays 1.0 until about +60 s, then grows to 13-30 ETH by +1200-1800 s (price x100-800 vs birth); LP gone to 0 between +600 s (ROTEO) and +3000 s (QI, Ollie, DOPE). Nobody can sell after the pull, so the fixed +350 s exit is the whole protection.
- Sellability by outside wallets (Swap logs of DOPE, +300..+800 s): 526 swaps, 374 buys, 152 sells from 148 distinct recipient wallets, all executed. So the token is not a honeypot for ordinary wallets. Still missing: our own $5 round trip (eth_call from a funded test address) before anything live.

## Why the "scripted crowd" objection does not kill it
The audit found a fixed wallet set (e71a first swapper in 73 of 74 pairs, 201 wallets in >=20 pairs). That means buyers are a repeating crowd that arrives after +60 s. We buy at +8 s, before any of them, at price 1.0, and sell at +350 s while hundreds of wallets are still buying. We do not need the crowd to be organic; we need it to show up on time (72 of 72 pairs did) and the LP to stay till +350 s (66 of 72 did).

## Money (estimate, not P&L)
About 4 pairs a day x (capped mean 7.5 - 1) x $10 = $260 a day gross, about 0.1 ETH/day at $10 per pair, before fill haircut. Even at half, 20x the 0.005 ETH/day bar. Real size limit: entry into a 1 ETH pool; $10-50 is fine (price impact under 5%).

## Ask
1. Testmgr/manager: reopen as a new arm "A2-DEC9R": same kit filter, buy +8 s (first Sync >= +5 s, 3 s lag), sell all at +350 s or WETH reserve < 50% of max. Record per trade: our fill, reserve at entry, reserve at +350, pull time if any. Kill rule: mean < 2x over 30 trades or 10 losses in a row; fast-track audit at 15 forward (we already have 8).
2. Exec/ops: $5 round-trip sell test on the next kit pair (eth_call sim first), no live switch (Ben only).
3. Old retirement reason ("not tradable") to be revisited: the forward n=8 and the 72-pair replay were never contradicted, only the buyer composition.
Caveats: replay in-sample on tape (21 Sep-9 Oct) but rule fixed a priori by analyst1 (#196) and forward n=8 agrees; the pulls are an operator event we cannot predict; if the operator changes the kit (decimals, seed), the arm dies silently.