# A1.6 How big can our buy be before a crew pulls? (Analyst 1, 6 Oct 2026, 13:00 Bangkok)

## Read this first
- **CORRECTION (A1.9, 13:50 Bangkok): the `0x7028bab73f82dff14b199593abb76c8b6acb3aa9` and `0x2dacdf0f44d07234832799a10a7cc5c8eff0223a` lines below are withdrawn.** Their "outside" early buyers were 20 crew wallets present in every coin, and "+464%" was my paper exit, not their real result (real: −24%). See http://159.223.88.168/hx-c9ois/sw/a1-9-non-pulling-crews.md
- **In crews A and B/C there is no bigger safe size. The ticket should shrink, not grow.**
  - Early outside buys of $9 or more (0.0035 ETH+, first 120 s) were followed by a pull within 60 s **48% of the time vs 14% expected** (47 pulls where 13.4 were expected, 98 buys).
  - The first buy of a wallet new to the crew: **71%** pulled within 60 s.
  - The bigger the buy, the worse: $9 to $20 2.8x the normal pull rate, $20 to $40 2.1x, $40 to $110 4.3x, $110+ 5.5x.
  - Only dust is let in. Buys under $9 are 1.25x normal. They come from bots buying $0.05 to $3 in 100+ crew coins each.
  - Real outside buyers of $9+ in the first minute of hot A or B/C coins: **lost 63% a trade** with the pull-time sell plan, **lost 11%** selling all at 2x (18 trades, 17 coins). The 9 that bought within 20 s made +54% at 2x.
- **Biggest size that doesn't clearly raise the pull rate in A and B/C: about 0.0015 ETH ($4).** Real buyers at $4 to $9 already averaged −13% (21 trades).
- **Crew `0x7028bab73f82dff14b199593abb76c8b6acb3aa9` is the opposite.**
  - 179 outside buys of $9 to $270+ in its first minute, 0 pulled within 60 s.
  - Those real buyers made **+464% a trade** (33 coins, 5 days).
  - Size can grow there; only depth limits it (see A1.7).
  - `0x2dacdf0f44d07234832799a10a7cc5c8eff0223a` is the same: 19 buys, 0 pulls, +148% (7 coins, 2 days).

| Crew | Outside buy size (first 120 s, all wallets) | Buys | Pulled within 60 s | Expected (same age) | Times normal |
|---|---|---|---|---|---|
| A + B/C | under $9 | 3,654 | 15.0% | 12.0% | 1.25x |
| A + B/C | $9 to $20 | 49 | 57% | 15% | 3.9x |
| A + B/C | $20 to $40 | 24 | 25% | 12% | 2.1x |
| A + B/C | $40 to $110 | 23 | 65% | 15% | 4.3x |
| A + B/C | $110+ | 20 | 65% | 12% | 5.5x |
| `0x7028bab73f82dff14b199593abb76c8b6acb3aa9` | $9 to $270+ | 456 | 0% | 0.1% | none |

## Warning on one wallet
- `0x90e2c652d8fe296dcb7fa38dff038c781fda2e23` made 18 of the $9 to $20 buys (2 to 4 Oct, about $10 each), and 83% were pulled.
- If that is Price Predict's own test wallet, the crews may know it. Without it, $9 to $20 is still 2.8x normal and all $9+ together 3.5x, so the result holds.
- Price Predict, please confirm whose wallet this is.

## Method
- **Data:** 1,893 pond coins from 20 Sep to 4 Oct and 1,114,613 swaps, from the droplet (pond tape + ffund first funders).
- **Army wallets:**
  - 12,963 wallets first funded by a crew distributor (a funder of 10+ wallets that trade 80%+ in one crew), or that trade 5+ coins of one crew at 90%+ share.
  - The known distributors come out on their own: `0xbae538011874261e3fb4ca6aa03cacb2766c4d81` for A and `0x79411dbcbcf0e5b27cd222f24e082fc0bbf32f16` for B/C.
  - "Outside" means everything else, minus the creator.
- **Expected rate:** for a buy at age t, the share of the same crew's other coins that were alive at t and pulled in t to t+60 s. This is fairer than a random 60 s window, because pulls bunch at certain ages. The random window is 3.5% (A) and 4.9% (B/C).
- **Caveat:** this shows that pulls follow outside buys, not for sure that the buys cause them. But the size ladder, the new-wallet 71%, and the fact that real buyers lost all point the same way.

## What to do
- **A and B/C:** cap the ticket at 0.0015 ETH ($4) in paper, or sell all at 2x (see A1.7). Never above $20.
- **`0x7028bab73f82dff14b199593abb76c8b6acb3aa9` and `0x2dacdf0f44d07234832799a10a7cc5c8eff0223a`:** $80 to $150 is fine; depth costs about 25 points at $150.
- Recheck weekly. Crews can change their pull trigger once they notice us.

First principles score: 85%. The time-matched rate and the real-buyer outcomes are as close to the source as on-chain data goes. Missing: proof of cause (only a live test with our own fresh wallets can give that), and the A/BC $9+ sample is small (98 buys, 18 real trades in hot coins).

Scripts: /root/a1/a16/army.py, a16.py, a16b.py; /root/a1/a17/nat.py (review container). Latest: http://159.223.88.168/hx-c9ois/sw/a1-6-buy-size.md