# Q-FLOAT and T4-ALIVE (DC Analyst 1)

## READ THIS FIRST
- **v1, 7 Oct 11:20 Bangkok. Both tests are KILLED by their pre-registered bars** (dc-prereg.md lines "Q-FLOAT" and "T4-ALIVE"; the T4-ALIVE line carries a time-fix line, because it was appended at 11:13 although it says 11:22).
- **Q-FLOAT (lever 3, who holds the float).** $500k crossers whose pool still stands at cross + 1 h: 160 trades (133 A, 27 B), $100, sell at 3x or 48 h, pull = 0.
  - Buy all: -$506 / -$158 a day. Winners: 10 / 3.
  - Cut chosen on half A: insider share in the bottom quarter. Result: -$69 / -$113 a day. Beats random in B only 11%.
  - Insider share as a predictor flips between halves: AUC 0.62 (A), 0.35 (B).
  - Outside net buy predicts nothing: AUC 0.49 (A), 0.36 (B).
  - Half B has only 3 winners, so it is also "too small".
- **T4-ALIVE** (T4-2h, plus price still at or above 50% of the high, plus net buying in the last hour): 126 / 60 trades, 25 / 11 winners. -$253 / -$277 a day (all T4-2h: +$526 / -$1,105). Beats random in B 54%.
- **What the data says, from first principles:**
  - At cross + 1 h, the launch bundle still holds a median **70%** of all wallet-held supply. The bundle is a median 93 wallets (every receiver in the first 30 blocks, plus the creator, plus wallets they pass tokens to).
  - After the coin first crosses $500k, **55% of those pools are still pulled within 48 h**, even though they stood at +1 h.
  - Neither a low bundle share nor outside buying tells which coins survive. At this stage, the operator decides the exit, and the chain does not show his plan ahead of time.
- **Push back, the bigger point:**
  - On my three tests today (pull gone, still alive, float share), every Lane D base loses money buying new coins hours after launch.
  - The money on this chain is made **before** the operator's exit: the First Window in the first minutes, the Second Window at the $1M gate.
  - So I suggest Lane D stop hunting new entries and do **Q5 HOLD MORE / SELL** for coins the other windows already own. There, the deep check only needs to answer "is the operator still in", and there is no new entry risk.
- **Clock check (Supervisor item 3):** every time in DC1.1, DC1.2, Q-FLOAT and T4-ALIVE is chain time.
  - v4: block time from the pp3_tfix anchors, and bar times converted back to blocks.
  - v3: swap block time from v3hist and v3live.
  - Q-FLOAT: pp3 t_cross (anchor-fixed); blocks from liqlock btime.
  - No launch-file times are used anywhere. The Lane F mixed-clock bug does not apply.

## Q-FLOAT method (as pre-registered)
- **Coins:** pp3.db cross at the $500k level (PP: $20k bought at cap >= $500k), cross 15 Sep to 5 Oct.
- **Dropped:**
  - 400 pulled by D (pp3 series lp_removed_share >= 0.5 at slot 6, or my pull time);
  - 141 outside the window;
  - 19 without 48 h of data;
  - 12 without transfer logs;
  - 9 other.
- **Insiders:** the creator, plus every non-market receiver in the first 30 blocks after the first market transfer, plus any wallet that receives the token straight from an insider before D.
- **Market side:** the pool, PoolManager, and the router/hub list from pp3_series.py, plus the second fomo router 0x0da5899dbb36bd741885d4228f04cd9c6304fc56.
- **F1:** insider share of wallet-held supply at D.
- **F2:** outside net buy in ETH from cross to D, from the pool's own swaps (buyer = last non-market receiver in the tx; seller = first wallet sending to the market), x ETH/USD.
- **Outcome:** first trade at or after D, the same DC0 v2 pricer as DC1.2.
- **Cuts** were picked on half A (F1 median 0.70, F1 q25 0.0009, F2 median $298). Random test: 10,000 same-size draws from the B trades.
- **Difference from the Supervisor's wording:** "operator bundle" here is the first-30-block crowd, not disperse funding or the same funder root. Proving funding needs one RPC lookup per wallet (about 15k wallets), so it was not done in this pass. A2's parallel ruler covers the funder-root version: compare when it lands.

## Files
- Q-FLOAT per coin (F1, F2, insiders, entry, multiple):
  - http://159.223.88.168/hx-c9ois/dc/dc-a1-3-qfloat.csv.gz
- T4-ALIVE trades (all T4-2h, with the `alive` flag and the hour's net buy):
  - http://159.223.88.168/hx-c9ois/dc/dc-a1-4-alive.csv.gz
- Scripts (research-box /root/dca1/):
  - http://159.223.88.168/hx-c9ois/dc/dc-a1-3-qfloat.py
  - http://159.223.88.168/hx-c9ois/dc/dc-a1-4-alive.py

## Limits
- Q-FLOAT has 156 v4 coins but only 4 v3: the pp3 v3 crossings carry a placeholder launch time.
- Half B is small (27 trades).
- No public router filter.
- The Engine could not score this yet. Its tape stops about 30 h after launch, and it has no float feature. The per-coin CSV can be joined by the Engine v1.2 (due 8 Oct 10:00 Bangkok) for an official rerun.

Cost: 0 X credits, $0 API, research-box about 25 CPU minutes (pp3_raw reads are disk-bound).

Version: v1, 7 Oct 2026 11:20 Bangkok, DC Analyst 1. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a1-3-qfloat.md