# A1.19 Who buys the coins that get pulled (Analyst 1, 6 Oct 2026, 17:24 Bangkok)

## Read this first
- **Nothing splits. No route or funding source is more than 2x as common in pulls as in our 5x winners in both halves, in a way that fires on coins our gate lets in. Close it. No gate change.**
- **The 490 instant pulls look nothing like our coins.** That's why the gate already never lets them in:

| Pre-pull or pre-gate buy ETH | Instant pulls A / B | Gate pulls A / B | 5x winners A / B |
|---|---|---|---|
| Old wallets (had ETH before block 60,000,000, weeks old) | **84% / 55%** | 0% / 0% | 0% / 0% |
| Disperse `0x84157d3d24630c297849c84d49ae7f1805c1b60f` | 0% / 0% | 0% / 54% | 2% / 57% |
| Disperse `0x03802649e469ed5680c58dff9e65425e22d21cc5` | 0% / 0% | 37% / 28% | 28% / 0% |
| Direct transfer from a normal wallet | 1% / 24% | 63% / 3% | 10% / 0% |
| Other contract | 15% / 22% | 0% / 15% | 59% / 42% |
| Wallets funded less than 1 h before buying | 0% / 22% | 93% / 96% | 85% / 79% |

- **Instant pulls are a sniper game.**
  - 490 coins (194 in half A, 296 in half B), rebuilt from chain. They were pulled a median 19 s after the first swap, after a median 7 buys and 0.054 ETH. 159 ETH in all.
  - 1,166 buyer wallets. 195 repeat snipers (5 or more of these coins each) carry 61% of the buy ETH. The top one, `0x1a45411c635f06a563cc61c8e96e6333dbc2420a`, bought 232 of the 490.
  - Their routes are small sniper contracts. For example `0x65050a9b7e5075a2ba5ced7b1b64ee66262c40dc` (752 bytes) is in 63% / 39% of instant pull coins and **0% of our 70 gated coins**, before the gate.
  - The same 195 snipers also bought in 69 of our 70 gated coins. So their presence tells nothing.
- **The pull coins come from two launch routes:**
  - The plain Uniswap position manager `0x73991a25c818bf1f1128deaab1492d45638de0d3`: 441 of the 4,239 launches it made since 13 Sep.
  - Launcher contract `0x9b226f198ef0038d054be6265aa9f76c0d6e5e78` (9,582 bytes): **49 of its 63 launches were instant pulls, and 0 reached the gate.**
  - No Kit A or Kit B launch was an instant pull.
- **The 5 gate pulls look like our winners**: fresh disperse-funded wallets, the same routers.
  - The only gap above 2x in both halves is "direct transfer" (A 63% vs 10%, B 3% vs 0%). It is tiny in B, and as a skip it loses money (below).
  - Both half A gate pulls (`0xce565efdc8527d54a147195a9f46ac559b46b5c2`, `0xa5a04c83d8c9228e7540978c65049a3bc41e3924`) were launched through the same Kit A copy, `0xa8b647a678d4c3edeab101f3a4be2d9dba9a4819`. Kit A stopped on 1 Oct, so there is nothing to act on.
  - Note: `0xa5a04c83d8c9228e7540978c65049a3bc41e3924` was pulled 157 min in, which is outside 2 h. I kept it so all 5 pulls are counted.
- **Skip filters priced anyway, at $1k with A2.21 frictions** (book = +$2,733 / +$3,535 a day):

| Skip if share of sampled pre-gate buy ETH | Coins skipped (pulls in them) | $ a day A / B |
|---|---|---|
| direct transfer over 2% | 19 (3) | +$1,756 / +$3,368 |
| direct transfer over 10% | 18 (3) | +$1,756 / +$3,757 |
| disperse `0x03802649e469ed5680c58dff9e65425e22d21cc5` over 5% | 24 (3) | +$1,946 / +$4,039 |
| disperse `0x03802649e469ed5680c58dff9e65425e22d21cc5` over 20% | 19 (2) | +$2,342 / +$3,928 |

  - Every cut loses money in half A. A pull costs about $1,000 at $1k, but each skip also throws away winners.
- **First principles:** a pull is the dev taking back the pool. The instant-pull devs do it before any real money arrives, against old sniper wallets, so a $1M gate can never be reached. The devs who pull after our gate let the same fresh crowd buy as in our winners first. Nothing in who buys tells the two kinds apart. Only the dev's own choice does, and that is not on chain until the burn.

## Method
- **Instant pulls:** every v3 WETH 0.01% pool created since 13 Sep 00:00 Bangkok (4,645, from factory `0x1f7d7550b1b028f7571e69a784071f0205fd2efa`). I read every Swap, Mint and Burn in its first 6,000 blocks (10 min). Pull = cumulative burns reach 50% of minted liquidity. Instant pull = 5 or more buys and 0 sells before the pull. Result: 490, a median 19 s, which matches A3.2.
- **Routes:** each buy transaction's `to`. Each buyer's signer was traced to its first ETH (binary search on eth_getBalance from block 60,000,000, then that block's transactions). All 1,166 instant-pull signers were traced.
- **Gated coins:** the A1.18 data (all pre-gate buys, 1,398 sampled funded wallets), split into the 29 targets, 36 losses and 5 pulls from `sw-a2-16-wide-trades.csv`.
- **Halves:** A = before 28 Sep 00:00 Bangkok, B = from then.
- **Money:** A2.21 per-trade x at $1k (`sw-a2-21-trades.csv`), 14 and 9 days.

## Limits
- Only 5 gate pulls (2 in A, 3 in B). Any split on them is fragile. The skip test on all 70 trades is the real check.
- An old wallet is anything funded before block 60,000,000. I did not trace its first funder further back.

## Asks for Supervisor
- Close pulls as a buyer-side signal. No gate change.
- Optional for PP/Coder, tracking only: a counter of launches by `0x9b226f198ef0038d054be6265aa9f76c0d6e5e78` (78% instant pulls). It never reaches our gate, so it is a scam tell for any earlier window, not ours.

First principles score: 85%. Every pull, buy and first funding block was read from chain, and all 490 coins were rebuilt with the same count as A3. Missing: more gate pulls, and the funders behind the old sniper wallets.

Scripts: /root/a1/a119/ip.py, ipset.py, iptx.py, rmix.py, pres.py, ipfund.py, fmix.py, skip.py (review container). Latest: http://159.223.88.168/hx-c9ois/sw/a1-19-pull-source.md