# Pond small buy test plan (v2, 5 Oct 2026, 22:00 Bangkok)

## READ THIS FIRST (for the AI that runs the test)
- v2 change: Price Predict now tags every new launch with its crew (from the launch wallet, before the first trade). Use that tag. Buy only coins from crews with a proven paper edge: crew A 0xcb7277454416d90e7fe8799cd300744d5797de03, crews B/C (same operator) 0x5bd9c2026df3d994fe1a0f051ac5992dfee85999 and 0x64a20b2a553c12960847d49fc381741a981de231, plus any crew whose own graded history shows 10x reached in 20%+ of its coins from the 10 s price. Skip crew 0xae3d565892c79fa8fc9243b380283dd419e28ec9 (10x only 7%).
- v2 rule for ALL pickers: a coin where repeat wallets (wallets that trade 20%+ of that crew's coins) make 80%+ of trades is a WASH coin. Wash coins get only this small buy lane ($5 to $20, sell orders attached). Never put normal size into a wash coin, whichever picker chose it: outside buys of 0.1 ETH or more get pulled on within seconds.
- Wash is not only the pond: crews 0x0424cab581eb1a66da3323d0452cff0872b0fc1f and 0xbce31c2e43027d3c926d502ceb9846053fb68ed1 (fee 0.05% and others) also show 97% of trades from repeat wallets. Crew 0x05c8b2d29a7bc2a5e9584f128f2ac3f8a7737bce has a broader crowd (repeat wallets 22% of trades) but 87% of its top 60 buyers were first funded by one wallet 0xa7bf3946d0551af07864b54a0854e85a24aeaaac, so treat it as unproven.
- Goal: prove or kill one idea with real money. Buy a tiny amount (about $5 to $20) of every new "pond" coin on Robinhood Chain about 10 seconds after its first trade, and sell automatically at 3x and 10x.
- Why it might work: pond coins are run by crews who do about 98% of the trading with their own wallets. The price they paint is real (their ETH is in the pool), so a small outsider can sell into it. On paper, across 1,402 pond coins (20 Sep to 3 Oct), 58% reached 3x and 20% reached 10x from the 10 second price before the pull.
- Why it might fail: crews pull the pool within seconds when an outsider buys big (0.1 ETH or more: pull within 60 s in 75% to 100% of cases, vs 3% to 5% after their own buys). Real small outsiders lost 45% overall because they never sold. Nobody has tested fills, fees and crew reaction with a repeating wallet.
- Hard rules: never more than 0.0074 ETH (about $20) per coin. Never add to a position. Every buy carries its sell orders at buy time. Run the existing safety gate (/opt/fm_gate.py) before every buy; any BLOCK means skip.
- Ben places or approves trades; automated mode (GMGN_ALLOW_AUTOMATED_TRADES=1) only if Ben switches it on.
- Report in Bangkok time, plain words, short.

## 1. What a pond coin is (detection, no crew lookup needed)
- Uniswap v4 PoolManager 0x8366a39cc670b4001a1121b8f6a443a643e40951, Initialize event topic 0xdd466e674ea557f56295e2d0218a125ea4b4f0f6f3307b95f85e6110838d6438.
- Pond = fee 100 (0.01%), no hook (0x0000000000000000000000000000000000000000), quote is native ETH (0x0000000000000000000000000000000000000000) or WETH 0x0bd7d308f8e1639fab988df18a8011f41eacad73, start market cap about 10 ETH, liquidity added by the creator through PositionManager 0x58daec3116aae6d93017baaea7749052e8a04fa7.
- About 108 pond coins launch per day. 99% come from about 11 crews. 100% are pulled within 24 h (for the three biggest crews the median life is 5 to 10 minutes).
- Clock starts at the FIRST Swap event on the pool (Swap topic 0x40e9cecb9f5f1f1c5b9c97dec2917b7ee92e57ba5563708daca94dd84ad7112f). Buy as soon as possible after first trade + 10 s. Log the real second we filled.

## 2. Paper results this plan is built on (entry at first trade + 10 s, exit only by take profit, everything else counted as a total loss)
| Group | Coins | Reached 3x | Reached 10x | Paper return, sell all at 10x | Paper return, sell all at 3x |
|---|---|---|---|---|---|
| All pond | 1,402 | 58% | 20% | +103% | +75% |
| Crew A, root 0xcb7277454416d90e7fe8799cd300744d5797de03 | 281 | 65% | 33% | +227% | +95% |
| Crew B, root 0x5bd9c2026df3d994fe1a0f051ac5992dfee85999 | 200 | 59% | 26% | +155% | +77% |
| Crew C, root 0x64a20b2a553c12960847d49fc381741a981de231 | 136 | 68% | 26% | +157% | +103% |
| Weakest big crew, root 0xae3d565892c79fa8fc9243b380283dd419e28ec9 | 312 | 48% | 7% | minus 26% | +43% |
- Half at 3x, half at 10x (all pond): paper about +89% per coin. This is the ladder to test.
- 10x levels last: median 32 minutes at or above 10x, 10% of cases under 90 s. A resting sell order has time.
- Paper ignores: fees (GMGN about 1% each way, gas about $0.015 a trade), slippage, buying later than 10 s, and any crew reaction to us.

## 3. The order (gmgn-cli, robinhood)
- Buy: `gmgn-cli swap --chain robinhood` with amount 0.0018 ETH (phase 1) or 0.0074 ETH (phase 2).
- Attach at buy time with `--condition-orders` and `--sell-ratio-type hold_amount`:
  - profit_stop, price_scale 200 (= 3x), sell_ratio 50
  - profit_stop, price_scale 900 (= 10x), sell_ratio 100
- No stop loss (a pull is instant; a stop does nothing useful and costs fees).
- Confirm the input token for robinhood with `gmgn-cli order quote` before the first live swap (the round trips on 5 Oct used the GMGN router 0x65050a9b7e5075a2ba5ced7b1b64ee66262c40dc and worked).
- Run from firstmin-trader 178.128.20.136 once Ben has set that IP as Trusted IP on the GMGN key. Space GMGN calls; a 429 ban grows with every retry.

## 4. Phases
- Phase 0, paper live (no money, 2 hours): detect every pond launch live, record first trade time, the price we would have filled at (first print after +10 s and after our real latency), and whether 3x / 10x printed before the pull. Pass if the paper ladder is positive and detection lag is under 5 s.
- Phase 1, $5 (0.0018 ETH) on the next 10 pond coins in a row. No picking. Pass if every order filled, the sell orders fired when price crossed, and fees were under 10% of the stake.
- Phase 2, $20 (0.0074 ETH) on the next 30 pond coins in a row. Max total at risk $600. Pass if net result is positive after all fees.
- Buy every coin from the allowed crews in a row (no further picking). Log the crew tag from Price Predict on every row.

## 5. Stop rules (stop at once, report to Ben)
- Total loss reaches $150.
- Two pulls inside 15 s after OUR buy in any 10 coins (after their own buys the crews pull within 60 s only about 3% of the time; two fast pulls in ten means they react to us).
- A sell order shows as triggered but did not fill.
- Any GMGN honeypot flag or gate BLOCK on a pond coin (means the pattern changed).

## 6. Log one row per coin
- Bangkok time of launch, token address (full), pool id, creator, first trade time, our buy time, seconds after first trade, ETH spent, tokens got, fill price, fees, gas.
- Max multiple printed after our fill before the pull, whether 3x and 10x printed, what we sold and at what price, ETH back.
- Pull time, seconds between our buy and the pull, ETH in the pool at the pull.
- Crew tag afterwards (section 7: crew A, crew B/C, or "other").

## 7. How to tag the crew afterwards (no data files needed)
- For the first 5 buyers of a coin (Swap events, then tx "from"), find each wallet's first funding (first block where its balance turned above 0, then the sending tx or the internal call inside it).
- Funded through send to many contract 0x5b8d85ebabf17cf6b67bfa2fe6795623951cd70e or 0xf341c935599822f1bc1ec29e27606724fbbf6595 by 0xbae538011874261e3fb4ca6aa03cacb2766c4d81 = crew A army. By 0x79411dbcbcf0e5b27cd222f24e082fc0bbf32f16 = crew B/C army. Both distributors were still active on 5 Oct 20:24 Bangkok and now also send through 0xd6d35b56237a05afd4cd7e441226fefc3f78c4e6.
- Wallet lists (2,832 army wallets) and the 617 crew coins exist in the review session; ask Ben if you need them.
- Background: robinhood-crews-top-coins doc in the Pro Trading project (how the crews work, cost, profit, pull trigger).

## 8. What would kill the idea
- Our real fills come late. Late entry is weaker: at 30 s the crew A 10x hit rate falls from 33% to 23% (still positive); at 60 s, selling all at 10x falls to about +35% to +78% paper.
- Crews start pulling on small outside buys once one wallet wins repeatedly.
- Fees plus slippage on $20 eat more than about 30% of the paper edge.

Version: v2, 5 Oct 2026 22:00 Bangkok (v1 bought every pond coin; v2 uses the Price Predict crew tag and adds the wash rule). Latest version: http://159.223.88.168/hx-c9ois/brx/pond-smallbuy-test-plan.md