# AN5.3 exits where the pool cannot be pulled (DC Analyst 5)

## READ THIS FIRST
- 7 Oct 11:36 Bangkok, v1. Prereg line "7 Oct 11:19 | A5 | AN5.3" written before any result. Paper only.
- **Where it ran:** Ben's Mac (read only copy of the DC Engine tape, nice 19). research-box was out of memory at 11:20 (97 MB free, load 53); I stopped my job there before it started work so the OOM killer would not hit PP's liqbuild2.
- **First principles:** with the pull gone, the loss is the dump. Holding 48 h on these coins ends at a median **0.02x** (H48 -$355 \| -$252 a day). The operator sells his float into the locked pool. So the exit must leave before the dump, same as before the pull.
- **Main result (NP: LP in locker `0x267444d099b10fb5ed7c3cc7b7c767adca574952` / `0xeff166aaf189323c58dc27ed1206eb2c37faacdf` or burned; chain clock $250k crossing; fill +30 s; 45 trades after dedupe):**

| Exit ($100, $ a day) | A (21 to 27 Sep) | B (28 Sep to 3 Oct) | beat random A \| B |
|---|---|---|---|
| **TRI10: 50% trail after 1.5x OR insiders sold 10% (best on A)** | **-38** | **-0.3** | 90% \| **99.3%** |
| TRN15: trail OR outside net buy < 0 over 15 min | -67 | +1.5 | 90% \| 99.5% |
| T15 | -72 | -41 | 96% \| 98% |
| TR50 (A2's best) | -81 | -54 | 95% \| 99% |
| NET5 | -109 | -10 | 29% \| 96% |
| X3 | -187 | -140 | 75% \| 92% |
| X5 | -215 | -158 | 67% \| 96% |
| H48 | -355 | -252 | 1% \| 0% |

- **Verdict:**
  - **The seller exit TRI10 passes the prereg kill bar in B** (beats TR50 and X3, beats random 99.3%). **First time reading the seller beats price alone.** It works here because there is no pull to outrun: the dump is slower, and the insiders' sells show up first.
  - **But it is TOO SMALL and makes no money:** 4 coins hit 3x in each half (bar 10). Best line is about $0 a day. An exit cannot make a dumped coin pay.
  - Same at a 5 min fill: best on A is NET5 (-$6 \| +$31, beats random 98% in B). Same story: break even, too small.
  - NPP (adds other non creator holders, minus `0xd0fd42bd8374ac659d3d81242d032211c2481055`): TRI10 -$76 \| -$55, loses to TR50 in B (+$177). Seller exit fails there.
- **Killed by me (look ahead):** the "$100k crossing" variant looked strongly positive (X3 +$141 \| +$99). It is **not valid**: the engine tape only holds coins that later reach a sellable $250k (2,758 of 2,791). Any entry below $250k on this tape picks winners by their future. Dropped, not reported as a result.
- **Forward lines (no backtest money claimed):** TRI10 on locked or burned coins; NET5 on T4 (from AN5.1). Both belong in the Coder's shadow book.
- Cost: 0 X credits, $0 API, 0 RPC, 0.5 h, Mac 3 s runtime.

## What it means
- **Lever 3 (float) is now the binding one.** No pull removes the 0x from pulls, but the median coin still ends at 0.02x after 48 h. The next test is the entry: skip locked coins where the operator still holds the float at the crossing (A1 Q-FLOAT, A2 op_share), then rerun these exits on what is left.
- **Seller exits need full wallets.** In 11 of 45 trades the coin fell 83% to 99% and no signal fired at all (51% of swaps have no wallet). With the thirdwindow tail (full transfer chains) TRI10 may fire on those too.

## Method
- Same code as AN5.1 (an5exits.py) with A2's coin lists (dc-a2-4-laned, dc-a2-5-float), A2's halves, one trade per ticker, creator and funder root per day. Script: http://159.223.88.168/hx-c9ois/dc/a5/an5nopull.py (md5 in status).
- LP exit not run (liqlock not on the Mac); on locked pools it equals X5.
- 4 of 45 entries are at a $3k to $27k cap 30 s after a "$250k" print (TAKO, LOAN, A Meme Coin, CREDITS): flash prints in thin pools. Kept, flagged.
- 26 to 29 of 30 hold trades run off the tape (wash stops ~30 h): slow exits are valued at the last price. Fast exits (TRI10, NET, T15) are not affected.

## Files
- http://159.223.88.168/hx-c9ois/dc/a5/an5np-exitmap-d30.csv , an5np-exitmap-d300.csv
- Per trade: http://159.223.88.168/hx-c9ois/dc/a5/an5np-trades-NP_250k-d30.csv.gz (and the other universes, same folder)

Version: v1, 7 Oct 2026 11:36 Bangkok, DC Analyst 5. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a5-3-nopull.md