# DC1.2 T4 survivor baseline (DC Analyst 1)

## READ THIS FIRST
- **v1, 7 Oct 10:58 Bangkok. Result: every T4 variant LOSES as a buy-all baseline.** By the pre-registered bar (dc-prereg.md line "7 Oct 11:00 | A1 | DC1.2", appended at 10:52 real time, see its time fix line) all of them are KILLED as a Lane D base: $ a day is at or below 0 in at least one half.
  - **T4-2h** (sellable $250k, pool not pulled by +2 h): A +$526, B -$1,105 a day. Half A is one coin: without the top 2 trades, A is -$658. 66 / 36 winners (3x).
  - **T4-6h:** -$704 / -$799 a day.
  - **T4-LOCK** (v4 LP burned or not held by the creator, at the cross): -$476 / -$121.
  - **T4-PONS** (Pons graduated pool): -$157 / -$12.
  - **Widened** (T4-1h, T4-2h-100k): both worse.
  - **T3** (re-ignition): 10 trades in 21 days. Too small.
- **Why, from first principles:** "not pulled" does not mean "alive".
  - The operator has two exits: pull the LP, or sell his float into the pool.
  - Survivors mostly took the second exit. At the T4-6h decision, 44% of entries are under a $20k cap: the coin crossed $250k, then fell 92%+ with the pool still in place.
  - The pull filter removes one way to lose, not the reason coins die.
- **So the T4 survivor base needs a "still alive" leg.** That is lever 3 (float vs outside money), which is my next item (Q-FLOAT).
- **Proposal for a new pre-registered line, not tested here:** T4-ALIVE = T4-2h AND price at the decision >= 50% of the sellable high so far AND outside (non-bundle) net buying in the last hour > 0.
- **Rulers used:**
  - 48 h horizon, covered by data;
  - label = sellable 3x;
  - sell at 3x or 48 h, pull = 0;
  - one trade per cluster (symbol, creator, crew root) per day;
  - $100 at pool depth.
  - Public router check is NOT applied (needs PP DCPP.3); all sells count.
- **Correction:** the first run (10:53) showed T4-2h half A at +$7 x 10^24 a day. A broken v4 bar close was used as the 48 h exit price. Fixed: the 48 h exit now uses the last SELLABLE price. All numbers here come from the fixed run.

## Results ($100 ticket, one per cluster per day; A = 15 to 28 Sep, 14 days; B = 29 Sep to 5 Oct, 6.4 days, v4 4.7 days)
| Variant | Venue | Trades a day | Trades A / B | Winners A / B | $ a day A / B | Without top 2, A / B | New $ (before $1M) A / B | Pull share |
|---|---|---|---|---|---|---|---|---|
| T4-2h | all | 32.4 | 431 / 231 | 66 / 36 | +526 / -1,105 | -658 / -1,190 | -358 / -799 | 28% |
| T4-2h | v4 | 16.2 | 248 / 56 | 30 / 5 | -280 / -492 | -477 / -589 | -81 / -381 | 40% |
| T4-2h | v3 | 15.1 | 183 / 126 | 36 / 21 | +806 / -686 | -302 / -771 | -278 / -468 | 21% |
| T4-2h | Pons grad | 2.4 | 0 / 49 | 0 / 10 | 0 / -57 | 0 / -131 | 0 / -50 | 0% |
| T4-6h | all | 25.0 | 340 / 172 | 23 / 22 | -704 / -799 | -897 / -904 | -416 / -446 | 16% |
| T4-1h | all | 39.6 | 535 / 275 | 88 / 46 | -917 / -1,397 | -1,124 / -1,473 | -727 / -1,432 | 38% |
| T4-2h-100k | all | 53.0 | 711 / 373 | 74 / 53 | -988 / -1,987 | -1,259 / -2,062 | -781 / -1,712 | 24% |
| T4-LOCK | v4 | 8.1 | 130 / 21 | 15 / 3 | -476 / -164 | -518 / -271 | -380 / -80 | 30% |
| T4-PONS | all | 4.7 | 47 / 50 | 6 / 11 | -157 / -12 | -197 / -102 | -157 / -8 | 0% |
| T3 | all | 0.5 | 6 / 4 | 2 / 0 | +7 / -27 | -21 / -24 | -2 / -27 | 0% |

- **Coins a day after cluster dedupe:** T4-2h 32, T4-6h 25. Dedupe removed 46 of 708 T4-2h trades (same symbol, creator or crew root on the same day).
- **The T4-2h half A outlier:** token 0xcc4cb99ef15e9799e1d7e45058683a72daf5b022 (v3), entered at a $6,298 cap on 22 Sep 19:04, first sellable print above 3x was 142.5x. Runner-up: 0x4ffefdfefc16daac253140125f50d8be9baffa52 (v4), 25.3x.
- **Pons graduation pools never pulled in this data**, but they still lose: graduation is not demand.

## Method
- Same coins and events as DC1.1 (research-box /root/dca1/ev_v4.pkl, ev_v3.pkl).
- **LP class (T4-LOCK):** liqlock.db liq + nft at the cross time, the same rules as A2 Q2: BURN if every open LP NFT sits at 0x0 or dead; CREATOR if any is held by the creator; else OTHER, which includes liquidity added by a non-PositionManager contract.
- **Crew root:** crew.db launch.root, or wallet.hop_root of the creator.
- **Data ends:** v4 at 5 Oct 17:39 Bangkok (liqlock research copy ends there, so later v4 pulls are unseen); v3 and Pons at 7 Oct 10:34.
  - A coin is dropped if entry + 48 h passes its venue's end.
  - Note for DC1.1: its v4 half B also relied on this liqlock copy, so v4 pulls after 5 Oct 17:39 are missing there too.
- **Script:** http://159.223.88.168/hx-c9ois/dc/dc-a1-2-t4.py
- **Every trade:** http://159.223.88.168/hx-c9ois/dc/dc-a1-2-trades.csv.gz (md5 4339b66e798d91704ce81df4c2dcc922).

## Limits
- No public router filter (PP DCPP.3 pending).
- Pons curve phase and Uniswap v2 not covered (same as DC1.1).
- v4 half B is small: 56 T4-2h trades.

## Asks for Supervisor
1. Accept the KILL of every T4 variant as a buy-all base. Lane D needs a "still alive" leg.
2. OK to pre-register **T4-ALIVE** (T4-2h + price >= 50% of the sellable high so far + outside net buy over the last hour > 0) and test it inside Q-FLOAT?
3. PP: the liqlock copy on research-box ends 5 Oct 17:39 Bangkok. Please refresh it, so v4 pulls after that are seen.

Cost: 0 X credits, $0 API, research-box about 6 CPU minutes.

Version: v1, 7 Oct 2026 10:58 Bangkok, DC Analyst 1. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a1-2-t4.md