# AN4.4 Funder roots for Uniswap v3 creators, and the v3 root lift (DC Analyst 4)

## READ THIS FIRST
- **v2, 7 Oct 13:32 Bangkok: LEAK CHECK PASSED. The rolling operator record survives with no look-ahead.**
  - Rule now: at each coin's FIRST TRADE (the earliest possible entry), look at the same operator cluster's coins launched in the prior 3 days. A coin counts as a **hit** only if its honest 5x already printed before that moment; as a **miss** only if it was already dead before that moment (pulled, or its 7 day window closed without a 5x); still open = **unknown**, left out. 15% of prior coins were unknown.

| Leak-free record (resolved prior coins only) | Half A: coins, honest 5x, lift | Half B: coins, honest 5x, lift |
|---|---|---|
| HOT: 3+ resolved, 2/3+ hit | 38, **61%** vs 24%, **2.56x** | 43, **63%** vs 33%, **1.89x** |
| COLD: 3+ resolved, 0 hit | 49, **12%** vs 28%, **0.44x** | 33, **18%** vs 39%, **0.47x** |
| (looser) HOT 2+ resolved, all hit | 17, 53%, 2.08x | 46, 63%, 1.91x |
| (looser) COLD 2+ resolved, 0 hit | 76, 15%, 0.51x | 57, 25%, 0.63x |

  - Still found after looking (not pre-registered as a coin rule); next it becomes E-AN4-5, a HOLD / SELL overlay on the books, pre-registered before scoring.
  - Remaining small leak to know: clusters join funders that fund fewer than 50 wallets in the whole table (a count over all days). It changes which wallets are merged, not any outcome.
- Updated per-coin file (columns kw, kl, ku = resolved hits, misses, unknown): http://159.223.88.168/hx-c9ois/dc/dc-an4-4-v3coins.csv

## v1 (13:31)
- **v1, 7 Oct 13:31 Bangkok.**
- **Built: roots for all 909 v3 coins** (896 DC1.1 universe v3 coins + 13 more book coins). Every FW mc5 and SW live book coin now has a root (book coverage was 4 of 71 and 12 of 177).
  - Dataset (owner DC Analyst 4, read only for all): research-box **/opt/tw/an4/crew_v3.db**, tables `wallet` (crew.db schema: addr, kind, funder, via, fund_b, fund_tx, nonce, src) and `launch` (pool, token, creator, root, hops, stop, b, src). PP's crew.db is not touched.
  - Method: PP ffund.py logic (archive balance gallop + debug_traceBlockByNumber callTracer; send-to-many contracts resolve to their caller), walked up until a contract, an exchange-like wallet (nonce ≥ 2000, the crew.db exchange floor) or 5 hops. 78,261 paid dRPC calls, 129 s, 0 errors.
- **The fixed list from half A does NOT carry to half B on v3** (prereg "7 Oct 13:28 | A4 | AN4.4"). Roots picked on half A have only **2 coins in half B** (1 hit). No lift can be measured: v3 operators do not keep the same funding chain week to week.
  - Even when chains are merged into operator clusters (369 clusters, 676 coins in clusters of 2+), only 7 clusters have coins in both halves (10 coins in B).
  - Inside a week the operators clearly differ: e.g. one cluster 12 of 17 coins honest 5x, another 9 of 13, others 0 of 14 and 0 of 11 (v3 base 26% in A).
- **New lead, NOT pre-registered (found after looking, hypothesis only): a ROLLING 3-day operator record works in both halves.**

| Operator cluster's last 3 days (only 5x hits that happened before this launch count) | Half A: coins, honest 5x | Half B: coins, honest 5x |
|---|---|---|
| 3+ prior coins and 2/3+ of them hit 5x ("hot") | 30, **63%** (rest 24%) | 29, **55%** (rest 35%) |
| 3+ prior coins and 0 hit 5x ("cold") | 56, **16%** (rest 28%) | 38, **16%** (rest 39%) |

  - Lift: hot 2.6x (A) and 1.6x (B); cold is a skip sign, 0.6x and 0.4x.
  - This fits first principles: an operator who just ran 3 coins up is still running the same play this week; a fixed list goes stale because funding chains rotate.
  - **Proposed next (prereg before any new read):** E-AN4-5 = hot / cold operator flag on v3, rolling 3 days, scored as a tilt and a skip on the FW mc5 and SW live books (their lab.py trades), and forward in Coder shadow.
- Cost: 0 X credits, paid dRPC 78,261 calls (12,822 requests), research-box 129 s, my container 1 min, 0.3 h.

## Method notes
- Coins and labels: DC1.1 universe (venue v3, honest 5x). Creator and pool creation block: harness gold idx.json.
- Clusters: union of creator→funder links (EOA to EOA only, funders that fund 50+ wallets in this table are treated as services and not joined). Picked by hand before the rolling test; the 50 cut and the 3-day window were not tuned.
- Chains: 795 of 909 hit the 5-hop cap, so the raw `root` column is the 5th hop, not a converged root. Use the cluster id (dc-an4-4-v3coins.csv `cl`) for operator identity.
- Pons graduated coins are not covered: their token creator is still missing (PP pons.db ask).

## Files
- Per coin (pool, token, creator, root, cluster, honest 5x, half, rolling record): http://159.223.88.168/hx-c9ois/dc/dc-an4-4-v3coins.csv (md5 4d173655626ce6bb3576a85c80bf08f7)
- Scripts: http://159.223.88.168/hx-c9ois/dc/dc-an4-4-roots.py (builder), http://159.223.88.168/hx-c9ois/dc/dc-an4-4-lift.py, http://159.223.88.168/hx-c9ois/dc/dc-an4-4-cluster.py

Version: v2 (leak-free record), 7 Oct 2026 13:32 Bangkok, DC Analyst 4. Latest: http://159.223.88.168/hx-c9ois/dc/dc-an4-4-v3roots.md