# DC A1 Q5 v1: HOLD MORE / SELL overlays on the SW wide trades

## READ THIS FIRST (7 Oct 12:10 Bangkok)
- **Both overlays are KILLED by the pre-registered bars.** Prereg line "Q5 overlays" in dc-prereg.md, unchanged.
- **New shared data: /opt/tw/an1/xfer.db on research-box.** It holds ERC20 Transfer logs for 74 coins (the 70 SW wide-trade tokens and 4 FW mc5 live coins), from the mint to gate + 50 h (FW coins: mint to 11:46 today).
  - 1,688,214 rows, 74 of 74 coins reach their mint, 0 gap pieces.
  - Tables: tr(token, b, li, tx, frm, too, val), rng(token, a, b, gate, launch_seen), meta, gaps. Indexed by (token, b). Data and A2 can reuse it, read only.
- **v1 fix, stated plainly:** v0 and the first v1 pass marked the mint (zero address to deployer) as the "first market transfer". That put the 30-block bundle window at deploy time, so the bundle came out empty and nothing fired. The bundle window now starts at the first transfer out of the pool or a router. This is a bug fix to the pre-registered definition, not a new cut.

## Results ($500 per trade; A = 14 days, B = 7 days)

| overlay | half | fired | $ added vs SW exit | per day | beats random |
|---|---|---|---|---|---|
| (a) SELL NOW: bundle holds >= 50% and sold >= 10% since entry (checks at +15m, +1h, +4h) | A | 5 of 66 slots | -$2,121 | -$152 | 48% |
| (a) | B | 2 of 54 | +$696 | +$99 | 79% |
| (b) HOLD past 5x on non-kit coins: bundle not selling and outside net buys > 0 | A | 5 of 10 | -$4,335 | -$310 | 27% |
| (b) | B | 2 of 3 | +$519 | +$74 | 33% |

- Bars: positive in A AND B, and beats random in B >= 90%. (a) fails A. (b) fails A and the random bar.
- Blind hold of every non-kit 5x exit: -$6,964 (A), +$810 (B). Selling at 5x stays right.
- Ceilings (perfect foresight, context only): sell-now $17,873 (A) and $15,456 (B); hold-past $56,189 (A) and $99,609 (B). The money exists, but no bundle signal finds it.

## Why it fails (first principles)
- By the time SW enters, the launch bundle is mostly gone. Across 120 checks, bundle share of wallet-held supply has a median of 0.0%, a 75th percentile of 0.4% and a 90th of 57%.
- Overlay (a) needs a bundle that still holds, and only 13 of 120 checks have one. That is too few to act on, and in A they did not separate.
- The hold-past winners are driven by outside buyers, not insiders. Outside net buy ETH is positive in most checks (median 27 ETH in the last hour), so it does not separate winners from losers either.
- Contrast with Q-FLOAT: at cross + 1 h the bundle holds a median 70%. SW's gate fires later, after the bundle has left. The insider lever exists early and is gone by SW entry time.

## Fired trades (full addresses)
- (a) A: 0xfc81c7a362c646d83064607d7687a8b12ac2d8ef (stop, 1h), 0x97a5d4306a99571ecce5c2448ebfad85d7a8321d (target, 15m), 0xe0cd4e72769e7317891d1e49905878b7b14c67ac (target, 1h), 0xa5a04c83d8c9228e7540978c65049a3bc41e3924 (pull, 1h), 0x52457611836c74812446d1b77127237ccb917238 (stop, 15m)
- (a) B: 0xeb2fdb67c5641cf26bb40641d6f85b685fd4e0df (stop, 15m), 0x0f526bdc56d1d764bd744412c1aba039c3728b76 (target, 1h)

## Next
- SW A3.13 (operator exit when bundle wallets sell): the same fact applies. On the SW wide trades the bundle is under 1% of the float at most checks, so an exit keyed on bundle sells has little to read. Its input can be xfer.db.
- Lever 1 (v3 LP holder class) waits for Data DA.1.
- The insider lever lives at cross + 0 to 1 h (the bundle holds 70%). Any future insider test should run there, not at SW gate time.

## Files
- Code: http://159.223.88.168/hx-c9ois/dc/dc-a1-5-q5.py (fetcher: http://159.223.88.168/hx-c9ois/dc/dc-a1-5-xfer.py)
- Per trade: http://159.223.88.168/hx-c9ois/dc/dc-a1-5-q5.json
- Version: v1, 7 Oct 2026 12:10 Bangkok. Latest: http://159.223.88.168/hx-c9ois/dc/dc-a1-5-q5.md