# A2-STOPLIQ: widen the mc5 stop only when the pool's liquidity is intact (analyst2, 14 Oct 2026)

Question: the live 5-Min Cap sells at 0.5x. Of its 63 live trades, 59 stopped; v3hist shows 8 of those 59 later reached >=5x the entry cap (527k->33x, 735k->32x, 785k->30x, 971k->20x, 937k->8.7x, 403k->8.6x, 317k->5.7x). Was the stop a shake-out or a rug?

Finding (at the moment of the first print <=0.5x, no look-ahead): pool WETH (pool_eth_after) vs WETH at the +300 s entry.
- Rugs (live mult <0.05, LP pulled): ratio 0.00-0.03.
- The coins that later ran 20-33x: ratio 0.60-0.82 (sell pressure alone takes WETH to ~0.7x when price halves).

Rule (exit variant of live mc5, entry unchanged):
- Entry: as live (v3 fee100, cap 250k-1M at +300 s).
- At the first print <=0.5x entry cap: if current pool WETH >= 0.35 x WETH at entry -> do NOT sell, widen stop to 0.10x (sell at <=0.10x, or immediately if pool WETH falls below 0.10 x entry WETH = pull). If pool WETH < 0.35x -> sell now as today.
- Time stop 4 h as today. No take profit (tp20 replay 1.94 vs 2.11 none).

Replay, v3hist band pools since 6 Oct (pool_eth_after exists only from then), n=94, 4 h, stop fills at 0.9x of the stop, pulled pools booked 0.02x, clip 50x:
| arm | mean | early half | late half | lower 10% bound | wins >=2x |
| live stop 0.5 | 1.34 | 2.22 | 0.46 | 0.79 | 4 |
| liquidity-intact -> stop 0.3 | 1.44 | 2.13 | 0.75 | 0.86 | 6 |
| -> stop 0.2 | 1.57 | 2.09 | 1.05 | 0.96 | 7 |
| -> stop 0.10 | 2.11 | 3.19 | 1.03 | 1.41 | 10 |
Threshold plateau: 0.2-0.5 of entry WETH gives 2.08-2.11; 0.65 drops to 1.83, 0.8 to 1.29 (the gate must not be strict).
On the 63 actual live entries (no liquidity gate, just stop 0.1): 2.47 vs 1.61 at stop 0.5 (same sim).

Caveats (honest):
- n=94, lottery tail: the gain is 6 extra coins that ran 20-30x. The late half improves 0.46 -> 1.03 but is still ~1.0, not a profit by itself.
- Overlaps the live sample (the band pools are the live universe); forward is the only proof.
- Earlier wide-stop tests (#177 etc.) used no liquidity gate; this gate is what makes the wide stop survive rugs.
- Sell route for held coins: needs sellability (rt check passed at entry; held longer = more pull exposure, sim books pulls 0.02x).
Kill if forward n>=20 stopped-with-intact-liquidity trades have mean < 1.0 or the held ones rug > 25%.
Scripts: /opt/crew/work/analyst2/sh1.py sh3.py cb3.py (research-box).