# Q2 Who can pull (DC Analyst 2)

## READ THIS FIRST
- 7 Oct 10:29 Bangkok, v1. Pre-registered as "Q2 who can pull" in dc-prereg.md (10:22) before any outcome was read. Same 612 T1 coins as DC2.1 (Uniswap v4, 21 Sep to 4 Oct). Paper only.
- **Who holds the LP NFT at the trigger tells you almost exactly who will pull.**

| LP control at trigger | Coins | Pulled within 24 h, A \| B | 5x hit rate, A \| B |
|---|---|---|---|
| CREATOR wallet holds an LP NFT | 538 (88%) | 98% \| 97% | 15% \| 18% |
| a CONTRACT holds it | 50 | 0% \| 11% | 6% \| 16% |
| BURN (all LP NFTs at 0x000000000000000000000000000000000000dead) | 20 | 0% \| 0% | 20% \| 10% |
| another plain wallet / none | 2 | | |

- **Pull bar: PASS.** The CREATOR class is 98 points (A) and 87 points (B) above the rest.
- **Money bar: a technical pass, but in substance a FAIL.** Skipping CREATOR coins makes the loss smaller in both halves: $100 tickets go from -$1,272 to -$236 a day (A) and from -$478 to -$58 (B). But that is only because the book shrinks. Per trade it is no better than random (beats random 30% in A, 50% in B), and no class makes money with a 5x-or-7-days exit.
- **What it means, from first principles:** a T1 coin whose creator holds the LP is a timed pump. The money is the spike before the pull. Locked or contract-held coins don't get pulled, but they don't run either.
- **The one useful link to DC2.1-F:** all 11 of its half A winners and 9 of its 11 half B winners are CREATOR-held coins. That rule works BECAUSE it catches the spike before the pull, not by avoiding pulls.
- **Use:**
  - (1) As an **exit tool**, not a skip: on a CREATOR-held coin, the pull is near certain (97 to 98% in 24 h, median 43 min after T1), so exits must be fast (sell at target, never hold).
  - (2) The 70 non-creator coins are the only T1 coins where HOLD longer is even possible. I suggest a follow-up test there with a slower exit (it is too small a sample, 41 in A and 31 in B, to call now).
- Per-coin classes: http://159.223.88.168/hx-c9ois/dc/a2/dc-a2-2-lpclass.csv.gz
- Cost: 0 X credits, $0 API, 33 small RPC calls (eth_getCode, batches of 20), research box read about 1 min, about 0.4 h.

## Method
- Research box liqlock.db: `liq` (every ModifyLiquidity with PositionManager tokenId) and `nft` (LP NFT transfers). Built 6 Oct 06:22. Our coins end 5 Oct, so the 6 Oct 13:14 liqlock gap does not touch them.
- At the trigger block: open positions = net liquidity per tokenId (PositionManager `0x58daec3116aae6d93017baaea7749052e8a04fa7`) or per non-PositionManager sender. Owner = the last NFT transfer before the trigger block.
- Class order, as pre-registered: BURN if every open NFT is at 0x0 or dead; else CREATOR if any NFT is held by the creator; else CONTRACT if any holder has code (EIP-7702 23-byte delegations count as plain wallets) or a non-PositionManager contract added the liquidity; else EOA.
- Main contract holders: `0x267444d099b10fb5ed7c3cc7b7c767adca574952` (1,969 bytes, 25 NFTs) and `0x23f8209572b4a1c2ad88a42749e830791fb027f1` (10,822 bytes, 20 NFTs). Not yet identified (locker or launchpad).
- Pull = PP wash `pull_b` (the same event as liq pulls, checked: identical rates). Money = DC2.1 v2 trade model (first trade 30 s or more after T1, 5x or pull or 7 d, depth-priced).

## Limits
- v4 only. Class CONTRACT mixes lockers and launchpads; the two big holders need naming.
- Pull and class come from the same liqlock tables, but class uses only events before the trigger block.

Version: v1, 7 Oct 2026 7 Oct 10:29 Bangkok, DC Analyst 2. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a2-2-pull.md