# A1.25: mc5 dead coin exit (v21 task)
First Window Analyst 1, 6 Oct 2026, 18:06 Bangkok. Code: http://159.223.88.168/hx-c9ois/fw/a1code/v3dead.py . Full grid log: http://159.223.88.168/hx-c9ois/fw/a1code/dead.log

## Read this first
- **FAIL. No dead coin exit. Base stays: stop 0.5x, 4 h.**
- H1 pick = **sell at fill +10 min if price is under the fill price (X 1.0), no buyer test**: H1 +$620 vs +$605 a day. On H2 it LOSES: **+$1,188 vs +$1,550 a day at $100** ($9,572 vs $12,658 at $1k), worst 3 days +$531 vs +$715, without top 2 +$620 vs +$894 a day.
- It cuts 29 H2 losers (mean loser 0.63 -> 0.51 ticket) but kills 6 H2 winners, incl. pool 0x6e8a6a1af8ec53f3f09c7303e5403558dee42d0a (35.88x -> 0.91x) and pool 0xcdc10abe9585ce4127ac4f4c1fe17daf34b0be7a (15.20x -> 0.80x). Winners on this rule often sit under the fill price at +10 min before they run.
- **The buyer test never bites:** new buyer wallets per 10 min are almost never 10 or fewer (median 472 new wallets in the first hour), even on coins that die. Fresh disperse-funded wallets keep arriving on dead coins too (A1.18), so "no new buyers" is not a dead signal here.
- Why the cheap losers can't be cut this way: the deep fills (A3: 19 of 116 stops under 0.3x) are pulls and one-block dumps. They happen fast and without warning; a slow-death coin is already caught by the 0.5x stop near 0.5x.
- Best H2-only cell (not pickable, H1 flat): +30 min, X 1.0, no buyer test: +$1,567 vs +$1,550 a day (+$17), worst 3 days +$843 vs +$715. Too small to act on; it is noise after a 48-cell grid.

## Setup
- Book: 178 dust-guarded mc5 trades (H1 31, H2 147), +5 min $250k to $1M, fill = pool state 6 s after the signal, $100 and $1k real depth (L from swaps), pool fee + GMGN 1% each way + $0.05 gas.
- Every exit (stop, dead exit) sold at the END OF THE NEXT BLOCK (pool state 0.15 s after the trigger), own buy kept in pool WETH; pull = 0; 4 h time stop.
- Dead exit: at T = +10/+20/+30/+60 min after the fill, if not already out: sell if price < X x fill price (X 0.7/0.85/1.0) AND new buyer wallets in the last 10 min <= K (K 0/3/10), or with no buyer test.
- New buyer wallet = first time an address receives the token (token Transfer logs, first swap to fill +61 min), excluding the pool and 1,459 router or bot addresses (100+ swaps as recipient or sender in the swap DBs).
- Pick on H1 by $ a day at $100; test once on H2. Pass = H2 $ a day up AND worst 3 days not worse AND $ a day without top 2 coins up.

## Base (next block exits)
| | trades | mean | $ a day $100 | $ a day $1k | worst 3 days ($100) | without top 2 ($100) | mean loser |
|---|---|---|---|---|---|---|---|
| H1 | 31 | 3.54x | +$605 | +$5,258 | -$357 | +$291 | -0.71 ticket |
| H2 | 147 | 2.27x | +$1,550 | +$12,658 | +$715 | +$894 | -0.63 ticket |

## Simple form (no buyer test), H2 at $100
| T | X 0.70 | X 0.85 | X 1.00 |
|---|---|---|---|
| +10 min | +$1,553 (3 cut) | +$1,463 (17 cut, 2 winners) | +$1,188 (35 cut, 6 winners) |
| +20 min | +$1,550 (0) | +$1,545 (6, 1 winner) | +$1,542 (14, 2 winners) |
| +30 min | +$1,550 (0) | +$1,543 (1, 1 winner) | +$1,567 (11, 2 winners) |
| +60 min | +$1,522 (1 winner) | +$1,520 (2, 1 winner) | +$1,501 (4, 3 winners) |
Base +$1,550. With a buyer test (K 0/3/10) almost nothing fires; every cell is within -$40 to +$15 of base.

## Winners killed by the H1 pick (+10 min, X 1.0), base -> dead exit
- 0xfafac5cd5f43916634ffc2a14607039a8c69ae60 2.98x -> 0.96x (24 Sep)
- 0xcdc10abe9585ce4127ac4f4c1fe17daf34b0be7a 15.20x -> 0.80x (25 Sep)
- 0x8a207f8a985b7996d9cd4af6469eb1768bc920c3 1.56x -> 0.86x (26 Sep)
- 0xd43679ab6a0d54401e2555fe77e1ca531b90f08b 3.29x -> 0.96x (26 Sep)
- 0x6e1a946e7a453d1c3afe054ff69024df4118e028 1.87x -> 0.75x (28 Sep)
- 0x6e8a6a1af8ec53f3f09c7303e5403558dee42d0a 35.88x -> 0.91x (2 Oct)
- 0xd305a2b6e1760c80ab06046985f86f3881c683aa 2.63x -> 0.86x (4 Oct)

## What would cut losers cheaper (proposal, not tested)
- The expensive losers are pulls and one-block dumps. The only lever left is sell speed (end of the same block as the first big sell, which live cannot do) or a pre-entry pull signal; A1.19 found no buyer-side pull signal and A2.19 none for cascades. So the 0.63 ticket mean loser is a fixed cost of the edge; wallet sizing should keep using it.

Version: v1, 6 Oct 2026 18:06 Bangkok. Latest: http://159.223.88.168/hx-c9ois/fw/a1-mc5-dead.md