# Second window: the v3 runner rule (6 Oct 2026, Bangkok)

## Read this first
- **Found a second-window rule that made money in both test halves.**
  - Coins on Uniswap v3: WETH pools with a 0.01% fee.
  - Buy when the coin first becomes sellable at $1M.
  - Only if it got there fast and the buyers look real.
  - Sell at 5x, or get out at a 40% drop.
- **Paper result, $500 a trade, honest exits:**
  - 59 trades in 17 days: +$26k, 13 of 17 days green.
  - Each $500 came back as $930 to $950 on average, in both halves.
  - It stays positive with the best 3 coins removed (1.72x of the stake).
  - 1 to 6 Oct alone: 20 trades, average 2.1x.
- **It is a lottery shape.** About 1 in 3 trades hits 5x. The rest mostly stop out at about 0.57x.
- **Not live.** Paper first. Ben says go.
- Pons graduated coins: no edge at any cap (0.7x to 1.0x on the first half). Plain v4: rugs (see the hold-or-sell doc).

## The rule (all four must be true)
1. **Venue:** Uniswap v3, quote WETH. Factory 0x1f7d7550b1b028f7571e69a784071f0205fd2efa.
2. **Gate:** the coin is sellable at $1M. That means buyers paid at least $20k at a market cap of $1M or more.
3. **Fast:** the gate came within 1.5 h of the pool's first trade.
4. **Real buyers:**
   - no single receiver took more than 8% of the buys;
   - almost none of the buying came from "army" wallets. An army wallet is one that bought early in 3 or more earlier coins.

**Then:**
- buy within 60 s (a 60 s delay still gave 1.75x on the second half);
- sell all at 5x;
- stop at a 40% drop;
- exit at once if half the pool's money is pulled;
- max hold 12 h.

## What we found (10 iterations)
1. **Where multi-million runners come from now.**
   - **Pons graduations** (879 new coins, 13 Sep to 6 Oct): only 5 reached sellable $5M and 1 reached $20M (AGE).
   - **v3 WETH 0.01% pools** (926 new coins): 60 reached sellable $5M and 24 reached $20M.
     - That's about 2 to 3 a day since 22 Sep.
     - The usual path: $1M in under 1 h, $5M at 2 to 7 h. Then the money is pulled 5 to 20 h after start (78% of these coins are pulled).
   - SC, SI, HI and VRAX (Oct runners) are all in this lane.
2. **Blind buy at the first sellable level, $500, sell 5x, stop 40%:**

   | Buy at | First half | Second half |
   |---|---|---|
   | $500k | 1.01x | 0.95x |
   | $1M | 1.44x | 1.30x |
   | $2M | 1.14x | 1.50x |

   Pons: 0.7x to 1.0x.
3. **Robustness of the blind $1M buy:**
   - delay 5 s to 15 min: 1.13x to 1.44x;
   - size $200 to $5k: 1.62x down to 1.08x (bigger buys move the price more).
4. **Signs at the gate, checked in both halves:**
   - fast to $1M is good;
   - one receiver taking many buys is bad;
   - a busy last 15 min is good.
5. **Fast plus spread-out receivers:** 1.75x and 1.78x (37 and 42 trades).
6. **Who launches these:** 878 different deployer wallets for 926 coins (fresh wallets every time). Deployer history can't be used. The LP is a v3 NFT the deployer can pull.
7. **Wallets before the gate** (1.7M transfers, 300 coins):
   - Army share high: bad in both halves (0.93x and 0.88x).
   - Lots of brand-new buyers: good.
   - Holder count and top-10 share: weak signal.
8. **All filters together:** 1.86x and 1.90x (28 and 31 trades), win rate 32% to 39%.
9. **Exit grid** (targets 2x to 10x, stops 30% to 50%):
   - all of them are above 1.19x;
   - 5x and 8x targets are best (1.76x to 1.83x);
   - small targets (2x, 3x) give up most of the edge.
10. **Daily money, $500 a trade:**
    - +$26k over 17 days, 13 days green;
    - worst day −$1.4k, best day +$5.2k;
    - at $2k a trade: 1.72x to 1.77x; at $5k: 1.50x to 1.55x.

## Honest limits
- The cut-offs were set on the first half (to 27 Sep) and tested on the second half (31 trades). Small sample.
- These are probably crew-made coins. Our 5x is sold into money that is in the pool, so that part is real. But the crews can change their game.
- Sell blocks inside the token contract were not tested on chain. Every coin did have outside sellers before the gate.
- Exits are priced from the pool's own depth after each trade. Each trade pays a 1% fee each way and a 5 s fill delay.

## What Price Predict must build (paper book `w2_v3_1m`)
1. A live listener for v3 WETH 0.01% pools: every Swap, Mint and Burn, plus token transfers for the first 2 h.
2. A live gate check:
   - sellable $1M ($20k of buys at $1M or more);
   - age 1.5 h or less;
   - top receiver share 8% or less;
   - army share near 0, from a wallet memory of early buyers in past coins.
3. Paper buy $500 at +5 s. Sell at 5x, stop at a 40% drop, pull guard, 12 h max. Price fills from the pool's depth.
4. A daily line in Bangkok time: trades, 5x hits, money, money without the best coin.
5. Go or kill after 14 days: go if positive without the best coin and 30+ trades.

## Receipts
- Data:
  - every Swap, Mint and Burn of 7,718 v3 WETH and USDG pools (13 Sep to 6 Oct);
  - every Swap of 2,243 Pons graduated pools;
  - token transfers before the gate for 300 v3 coins;
  - the Pons curve tape (pons.db), read only.
- Scripts, container /tmp/w2: v3/fv.py, a/fa.py, an/v3build.py, an/sim2.py, an/feat.py, an/wal.py, v3/mints.py.

Version 1, 6 Oct 2026 Bangkok. Latest: http://159.223.88.168/hx-c9ois/hx/second-window-v3.md