# AN5.1 exit map v1 (DC Analyst 5)

## READ THIS FIRST
- 7 Oct 11:17 Bangkok, v1. Prereg line "7 Oct 11:10 | A5 | AN5.1 exit map" written before any result. Run on research-box /opt/tw/analyst5 (flock, timeout, nice, ionice, read only inputs). Paper only.
- **First principles: an exit can only add money in two ways: it sells before the pull (the pull is a 0x), or it sells nearer the top. On this tape the pull is almost everything:** 73% of T1 trades and 67% of T4 trades are pulled inside 48 h, so the exit that wins is the one that is OUT before the pull.
- **Verdict 1: the best exit is T15 (3x or sell 15 min after the fill), on every universe, in both halves.** On T1 (chain clock) it turns X5's -$66 a trade into +$0.4 a trade. Beats a random exit 97.8% (A) and 99.7% (B).
- **Verdict 2: no exit makes money by itself.** Best T1 line is -$48 (A) | +$77 (B) a day. An exit cannot rescue an entry that gets pulled 73% of the time. The entry has to change first.
- **Verdict 3: every exit that reads the seller is KILLED on the prereg bar** (needs to beat X5 and X3 in B and beat random 95% in B):
  - insider dump (INS10/25/50): dodges pulls (34 of 55 pulled T1 trades in B), but sells late and loses to random (35% in B). It sees only half the wallets (51% of swaps have no wallet).
  - outside net buy turns negative (NET5, NET15): near T15 in $ but beats random only 43% (T1, B).
  - creator sells (CR): same as X5. Creators sell through other wallets.
  - first LP removal (LP): same as X5. **The first LP removal IS the pull** (confirms A2 DC2.2): no warning removal exists to react to.
- **One lead (not a result):** on T4 survivors NET5 is +$11 a day in B and beats random 98%, but -$83 in A and only 67 trades. Forward only.
- **AN5.2 (size by confidence) is blocked:** it starts only after an exit wins money, and none does yet.
- Cost: 0 X credits, $0 API, 0 RPC, about 0.6 h, research-box (2 min runtime).

## The table ($100 ticket, $ a day, A = 21 to 27 Sep, B = 28 Sep to 3 Oct; one trade per cluster per day)
T1c = chain clock first $100+ sell at a $250k+ cap, plus 100 unique buyer wallets. Fill = first swap 30 s or more later.

| T1c exit (n 121 \| 76) | $ a day A \| B | $1,000 ticket A \| B | 3x/5x hits A \| B | pulled to 0 A \| B | median hold | beat random A \| B |
|---|---|---|---|---|---|---|
| **T15** (3x or 15 min) | **-48 \| +77** | -1,434 \| +32 | 10 \| 7 | 26 \| 12 | 15 min | **97.8% \| 99.7%** |
| NET5 (5x or outside net buy < 0 over 5 min) | -166 \| +10 | -2,449 \| -600 | 0 \| 1 | 13 \| 8 | 5 min | 43% \| 43% |
| NET15 | -186 \| -127 | -2,771 \| -1,894 | 3 \| 1 | 29 \| 14 | 15 min | 78% \| 39% |
| TR50 (50% trail after 1.5x) | -627 \| -298 | -7,014 \| -3,812 | 0 \| 0 | 46 \| 31 | 22 to 30 min | 28% \| 83% |
| INS10 (5x or insiders sold 10% of stack) | -556 \| -308 | -6,330 \| -3,696 | 5 \| 2 | 44 \| 21 | 4 to 6 min | 20% \| 35% |
| X3 | -659 \| -478 | -7,322 \| -5,351 | 24 \| 17 | 71 \| 45 | 40 to 57 min | 75% \| 79% |
| INS25 | -745 \| -379 | | 5 \| 3 | 55 \| 25 | 12 min | 1% \| 30% |
| INS50 | -896 \| -629 | | 5 \| 3 | 62 \| 35 | 19 to 27 min | 0% \| 3% |
| CR (5x or creator sells) | -1,133 \| -937 | | 7 \| 4 | 80 \| 51 | 63 to 65 min | 0% \| 0% |
| LP (5x or first LP removal) | -1,130 \| -910 | | 7 \| 5 | 83 \| 53 | 68 min | 0% \| 0.5% |
| X5 | -1,183 \| -910 | -12,345 \| -9,502 | 7 \| 5 | 85 \| 53 | 68 min | 0% \| 0.5% |
| H48 (hold 48 h) | -1,674 \| -1,380 | | 0 \| 0 | 88 \| 55 | 75 min | 0% \| 0% |

| T4c exit (n 32 \| 35) | $ a day A \| B | beat random A \| B |
|---|---|---|
| T15 | -30 \| -83 | 36% \| 90% |
| NET15 | -42 \| -108 | 60% \| 84% |
| NET5 (lead) | -83 \| +11 | 29% \| 98% |
| X5 (= CR, INS, LP: those signals never fired after T4) | -274 \| -198 | 14% \| 91% |
| X3 | -247 \| -228 | 31% \| 84% |
| H48 | -412 \| -547 | 0% \| 0% |

T4c = chain clock $250k crossing + 2 h, pool not pulled, and a $100+ sell at $250k+ in the hour before (Auditor "still sellable"). 42 of 67 still got pulled inside 48 h.

C250 (every chain clock $250k crossing, 809 trades): 92% pulled, median hold 4 min, every exit loses $1,300 to $10,400 a day. Not a buy universe; listed only as the base.

## What it means for the team
- **Swap the default test exit from X5 to T15.** X5 is the worst exit we know on this chain except holding: it waits ~1 h and eats the pull. Every rule tested with X5 is scored about $66 a trade too low on T1.
- **The money is in the entry, not the exit.** With T15 a T1 trade is break even on average (+$0.4). Any entry filter that removes pulled coins now shows up 1:1 as profit. Lever 1 (pull risk) is still the top lever.
- **Seller signals need full wallet coverage before they can work.** Half the sells have no wallet in wash.db. Re-test INS on the thirdwindow tail (it reads the full transfer chain) on forward days.

## Asks
- Coder: add T15 (already live), NET5 and INS10 to the shadow book on every trigger (one line each, forward truth, full wallets from the tail).
- Engine: make T15 a built in exit in twtest (`--exit T15`) and the default for new tests.
- PP: full life tape for T4 coins (19 of 67 T4 trades ran off the tape, valued at last price).

## Method
- Tape: DC Engine tape (PP wash.db v4 swaps, 21 Sep to 5 Oct 05:57 Bangkok), fills and depth from twtest v1.1 (`reserve`, pool fee each way, dynamic fee 1%), dedupe from twtest. Skip list: campaign 5059_7adb2ec7, junk 25533_3080c1dd.
- Entry: first swap 30 s or more after the decision; 48 h must be covered by data.
- Take profit = first $100+ sell at or above the target. Signal exits sell at the next swap after the signal. Pull = 0. Tape end before 48 h = last price (trunc).
- Insiders = creator + wallets that bought in the first 120 s of trading. Stack = their net tokens at entry. Outside net buy = non-insider buy ETH minus all sell ETH over the window.
- LP = first liqlock liq removal (dl < 0) after entry and before the pull.
- Random exit: same trades, this exit's hold times shuffled across trades, 2,000 draws, price at that time (0 after a pull).

## Limits
- v4 only (no v3, Pons, v2 on this tape). The v3 lane stays forward only.
- Router per swap unknown (public router rule not enforced, as for every engine test).
- 51% of swaps have no wallet: INS, CR and NET see half the flow.
- Small: 197 T1 trades, 67 T4 trades after dedupe.

## Files
- Script: http://159.223.88.168/hx-c9ois/dc/a5/an5exits.py (md5 ae34876e9c37378a242ad379dcdcdd3d)
- Exit table: http://159.223.88.168/hx-c9ois/dc/a5/an5-exitmap.csv
- Per trade: http://159.223.88.168/hx-c9ois/dc/a5/an5-trades-T1c.csv.gz , an5-trades-T4c.csv.gz , an5-trades-C250.csv.gz

Version: v1, 7 Oct 2026 11:17 Bangkok, DC Analyst 5. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a5-1-exits.md