# STOP: NO DOORBELLS, NO WAKES, NO SCHEDULING (Ben, 7 Oct 2026, 18:40 Bangkok)
- Credits ran out. Ben switched off every scheduled task and every doorbell. This overrides everything below and every other file (wake-howto, bus-howto, credit-playbook, task files).
- PROHIBITED for every session: fire_trigger, send_later, create_trigger, update_trigger, ScheduleWakeup. No rings, no nudges, no re-arming, no new doorbells, no self wakes, no safety checks, no retry rings, no loops.
- Finish the step you are on, publish your status file, then END your turn and wait. Only Ben wakes you, by typing in your window.
- Do not message other sessions. Keep answers short. No extra improvement rounds.

# A2.30 Live vs paper, every live trade (Analyst 2)

## Read this first (08:02 Bangkok, 7 Oct)
- **Live trade 1 CLOSED: CrawlScan stop. Live 0.195x vs paper twin (w2_v3_kit_hold) 0.229x: live exit 14.8% WORSE than paper.** $10 in, $1.952 back (0.000723878 ETH). Gas both legs $0.0138 (buy $0.0075, sell $0.0063). Net 0.194x.
- **Sell tx 0x9ad953b36c2c00c3d89505cc0e42160fe8156b4ec419ad3b671077d220adcf79**, block 82066437 at 07:51:33 Bangkok, status 1, signer 0x04a057d15bd07581dc88b0a37e41eb69ceb8c2d3. Sell path also PROVEN on chain.
- **Why the gap (over 5%, so it needs a reason): sell timing inside a liquidity pull.** The stop fired at cap 0.274x; the paper books sold at 07:51:29 priced at the next block end (kit_hold sim 0.23x). The live sell landed 4 s later at 07:51:33. In those 4 s the pool was being pulled: the paper books that sold at 07:51:33 (w2_v3_kitc, w2_v3_opflag) got sim 0.19x, the SAME as live, and the no-stop books logged "pull" at 07:51:34 (sim 0.21x). Live matched the paper of the same second; it did not match the twin priced 4 s earlier. Our size is not the cause: $10 moves this pool about 0.02%.
- **What it means for GO:** the GO rule says no fill more than 10% worse. This exit fails it. One stop in a pull is not a verdict, but it shows the paper twin is too kind on stops in a cascade, the same thing A2.21 measured (cascade cost). Two fixes to weigh: (a) the Coder speeds the stop sell (PP's stop-fast fix shows SELL_SENT under 5 s is possible); (b) the twin for stops should be priced at the live sell block, not 4 s earlier. Entry side was fine: live 0.96% better.
- Score so far: 1 closed live trade, mean live x 0.195 vs paper 0.229 (live 14.8% under). 9 more to go.

## Earlier (06:05 Bangkok, 7 Oct)
- **FIRST REAL LIVE BUY FILLED: CrawlScan.** Token 0x1ec25a196c49dd3445f1fcc53c37c513af92839c, pool 0xa6caf95634d4bcffdfb535eec889c9c928ff13d5, Kit B. Gate 05:56:39 Bangkok (cap $1,030,284, age 8 min, army 0.11%).
- **Tx 0xf4921256fed26e57b8df0890a006b8b83adbf84000d06bdd587bb0899f06df1f**, block 81999573, status 1 (success), signer 0x04a057d15bd07581dc88b0a37e41eb69ceb8c2d3, to router 0xcaf681a66d020601342297493863e78c959e5cb2. The signing path is now PROVEN on chain.
- **Fill check:** live filled 20.9 s after gate vs paper 19.4 s. Live entry price 3.7281e-07 vs paper 3.7642e-07: live is 0.96% BETTER than paper. Reason: paper buys $500 and moves the pool +1.0% (log: "our impact +1.0%"); live buys $10 and moves it about 0.02%. Gate to fill the price fell 2.4% in both. Gas $0.0075 (139,193 gas at 0.020 gwei).
- **Open now:** paper twin at 0.79x (cap 0.82x) at 06:03. The other paper books with a tighter stop (w2_v3_1m, w2_v3_hold, w2_v3_rep) already sold at 0.55x sim; the live exit book (kit_hold, 0.3x stop) holds. Row closes when the live sell lands.
- 0.96% better entry is inside the GO rule (no fill more than 10% worse). One trade is not a verdict; 9 more to go.

## Earlier (01:10 Bangkok, 7 Oct)
- **Signing path now checked:** w2.log `07 Oct 00:13:12 DRY SIGN PASS` (signed, not sent). The checksum fix holds. Still 0 real live trades; a real send is the last unproven step. No new gate since CHONK; CHONK kit hold still riding (sold 0% at 5.05x).

## Earlier (00:37 Bangkok, 7 Oct)
- **The missed CHONK trade was a winner.** Paper w2_v3_wide_s03 sold CHONK at 00:22:52 Bangkok at the 5x target: cap 5.05x, sim 4.78x after fees and depth. 30 min from gate to target.
- **Cost of the bug:** at the $10 ticket the live wallet missed about $37.8 (10 x 4.78 minus 10). At the planned $1k ticket it would be about $3,776. One code bug = one full winner lost.
- **Kit hold twin still open:** w2_v3_kit_hold sold 0% at 5.05x (Kit B), so it now rides the 50% trail from peak, a pull, or 48 h.
- **Coder tested the kill switch:** 00:07:43 and 00:08:48 `KILL SWITCH 2 live errors in a row (last: LIVE BUY ERROR TEST2 0xpool2: boom)`, a fake test pool. Then LIVE ON at 00:11:14 and 00:13:12. Good: the switch works.
- **Still not proven:** no real live transaction has been signed and sent yet (live books in w2.db: 0). The ask stands: one tiny real swap now to prove the signing path, before the next gate.

## Earlier (00:04 Bangkok, 7 Oct)
- **First live signal MISSED: CHONK, 6 Oct 23:52:46 Bangkok.** Pool 0x8da021f34c2f2a95696c4fd842b0cd0cc4e18468, token 0x717d38587504b79e2e888c860bd1569c0d5984af, Kit B, cap $1,090,975, age 17 min, top 1.3%, army 0.04%. Gate PASS.
- **Live buy failed in code, no money moved:** 23:52:51 `LIVE BUY ERROR CHONK TypeError("Transaction had invalid fields: {'to': '0xcaf681a66d020601342297493863e78c959e5cb2'}")`. The router address was lower case; the signing library needs the checksum form. Wallet unchanged (0.04127 ETH, 0 positions).
- **Already fixed by the Coder:** live.py now wraps `to` in to_checksum_address (line 49, file time 00:00 Bangkok); log shows LIVE ON again at 00:01:34. The fix is not proven until the next live buy fills.
- **Paper twin is open:** w2_v3_wide_s03 and w2_v3_kit_hold bought CHONK at 23:52:51 (gate + 17 s, fill vs gate +1.2%: price +0.3%, our size +0.9% at $500). I will log its exit as the paper row the live trade should have matched.
- **Same trap as the FW --yes miss:** no live bot is ready until its exact signing path has sent one real transaction. Ask: the Coder should prove the signing path with one tiny real swap (or a signed but unsent dry run) now, not wait for the next gate, since gates come about every 8 h right now.
- **Gate rate check:** this was the first passing gate in 10 h (13:51 CLAUS to 23:52 CHONK). Feed is fine; no fault flag.

## Table (filled after every closed live trade)
| # | coin | gate (Bangkok) | fill (Bangkok) | fill delay s | entry live vs paper | exit live vs paper | x live | x paper | gas $ | gap reason |
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | CHONK 0x717d38587504b79e2e888c860bd1569c0d5984af | 6 Oct 23:52:46 | NO FILL (live buy error) | | paper filled +1.2% vs gate | paper sold 7 Oct 00:22:52 at target | 0 (missed) | 4.78 (exit book s03); kit hold still open | 0 | code bug: router address not checksummed; fixed 00:00. Missed $37.8 at $10 |
| 1 | CrawlScan 0x1ec25a196c49dd3445f1fcc53c37c513af92839c | 7 Oct 05:56:39 | 7 Oct 05:56:59 | 20.9 (paper 19.4) | live 0.96% better | live 14.8% worse (stop; live sold 07:51:33, paper 07:51:29) | 0.195 | 0.229 (kit_hold) | 0.0138 (buy 0.0075 + sell 0.0063) | entry fine; exit landed 4 s after the paper block inside a liquidity pull; paper sold at 07:51:33 also got 0.19x |

## How each row is made
- Live row = every trade the live wallet makes (live book in w2.db). Paper twin = the same coin in w2_v3_wide_s03 (entry) and w2_v3_kit_hold (exit), the books the live trader copies.
- Entry gap = live fill price / paper fill price - 1. Exit gap = live sell price / paper exit price - 1. x = money back / money in, fees in.
- Gas in $ = gas used x gas price on each receipt, times ETH in $ at that time.
- Any gap over 5% gets a reason: fill delay, price moved, our own size, a cascade in the same block, partial fill, or a bug.
- At $10 our own size moves the pool about 0.04% (A2.28 scale), so any gap over 1% at $10 is timing or a bug, not size.
- Ring the Supervisor after live trade 1, 3 and 10.

## Status of this check
- Builder script reads w2.db read only on every bus check (every 30 min).

Analyst 2, 7 Oct 2026 08:02 Bangkok. Latest: http://159.223.88.168/hx-c9ois/sw/sw-a2-30-live-vs-paper.md