# Hold or sell at $250k (phase 2, v1.3), 5 Oct 2026 Bangkok

## Answer in one line
For almost every coin, sell at $250k, and do it within 60 seconds. Hold only BUNDLE family coins, and only while they stay alive. Even there the edge shrank in the newer half, so test it on paper first.

## The data (read only, nothing written to PricePredict)
- 2,133 new Uniswap v4 coins that hit the gate (first $100+ sell at $250k+), 15 Sep to 2 Oct. Each one was followed 72 h or more.
- Every swap and every liquidity add or remove in each coin's pool, from the coin's start until 5 Oct. Pulled straight from the chain for these pools only, into a scratch folder.
- Why not PP chainswaps: on 5 Oct 16:15 its 12 shards were still running, about 10% done each. Each shard only had the first few hours of its slice, so paths after the gate had holes.
- Exit value is honest:
  - your sell moves the price (we use the pool's real depth at that moment);
  - if the liquidity is pulled, your coins are worth what is left, often $0;
  - 1% fee;
  - bag = $3,000 at the gate.
- Learn = gates up to 24 Sep (564 coins). Test = after (1,569 coins).

## What we found (5 iterations)
1. **Plain coins (no hook, 1,795): 95% get their liquidity pulled within 72 h.**
   - The pull comes a median of 5 min after the gate: 10% within 71 s, 25% within 3 min, 75% within 9 min.
   - Selling at the gate keeps 0.94 (learn) and 0.88 (test) of the bag's value.
   - Every hold rule keeps 0.11 to 0.43. Holding 72 h keeps about 0.
2. **Who pulls: the main launcher.**
   - The launcher is 0x58daec3116aae6d93017baaea7749052e8a04fa7 (1,762 of the gated coins).
   - Each coin's liquidity is an NFT owned by the creator, who burns it and takes the money out.
   - NFT sent to 0x000000000000000000000000000000000000dead before the gate: 18 coins, 0 pulled.
   - NFT not sent there: 1,744 coins, 1,726 pulled (99%).
3. **Even coins that cannot be pulled fade.**
   - For the 24 locked coins: sell at the gate 0.93, hold 72 h 0.04, best trailing stop 0.70.
   - Only 1 in 24 reached a sellable $1M.
   - The 88 plain coins never pulled: sell 0.88, best hold 0.88. No edge.
4. **BUNDLE family (121 coins, gate about 20 s after launch) is the only place where holding pays.**
   - Rule: sell all at a sellable $1.5M. Get out if the cap falls under 70% of the gate or the liquidity drops by half. Max 6 h.
   - Result: 1.42 vs 0.91 for selling at the gate.
   - But by half: learn 1.58, test 0.99 to 1.08 (selling = 0.92). The edge is shrinking.
   - The $1M target with a 50% stop: hold minus sell = +0.37 per $1 overall (90% range +0.21 to +0.53). Test half: about even.
5. **Stage checks after the gate (1 min, 5 min, 15 min, 1 h). Here "alive" means price at or above the gate AND liquidity at 80%+ of the gate.**
   - BUNDLE alive: holding beats selling at that point in both halves. Test half: 1 min 1.28 vs 1.11, 5 min 1.72 vs 1.40, 15 min 2.36 vs 2.07. At 1 h it is even.
   - Plain alive: selling wins every time. Test half: 5 min 1.37 vs 0.23; 1 h 2.06 vs 1.24.
   - Other hooks alive: selling wins.
   - BUNDLE not alive: worth about 0.05 to 0.38 by then. Sell at once.
6. **Speed matters for plain coins.** Selling at gate +60 s keeps 0.89. At +5 min it keeps 0.57 in the test half.
7. **Bag size.** Pool depth at the gate is a median $22k (plain), $33k (bundle), $60k (other hooks). A $3k bag loses about 10% to its own sell; a $1k bag about 4%. That loss is already in the numbers above.

## The process (v1, paper only)
1. **Second 0 (at the buy):** read the hook code of the pool.
   - Campaign (5059 bytes): never buy.
   - BUNDLE (341 bytes): quick lane.
   - Anything else: sell lane.
   - Main launcher only: check whether the LP NFT is at the dead address (log it; it does not change the call yet).
2. **At $250k (the gate):**
   - Sell lane: **sell 100% within 60 s.** No exceptions found in 2,012 coins.
   - Quick lane (BUNDLE): **hold.** Sell all at a sellable $1.5M. Get out at once if the cap falls under 70% of the gate or the liquidity drops by half. Max 6 h.
3. **Quick lane checks at 1, 5 and 15 min:** keep holding only if it is alive (price at or above the gate and liquidity at 80%+). If not, sell now.
4. **Weekly:** redo the test on the newest week. Drop the BUNDLE hold if it loses to selling at the gate on new weeks.

## What this means for "keep some for later"
- On the v4 launch tape there is nothing to keep. The money in a slow runner was not in these coins.
- The 107 runners since July mostly traded on other venues (Pons V2, v3, pools.trade).
- So the hold question for slow runners needs those venues. That is the next test once PP chainswaps finishes.

## What I need from PricePredict (PP builds, I only read)
1. **Finish chainswaps.** On 5 Oct 16:15 every shard was about 10% done. When finished, I redo this test on all venues.
2. **A liquidity table per pool:** every add and remove with time, the share of the pool removed, who did it (v4 sender + salt), and the pool's depth (USD) after.
3. **An LP lock table:**
   - main launcher: the NFT id per coin, its owner over time, and the time it went to the dead address;
   - v3 / v2: LP token burns and lockers.
4. **Hook family per pool** (code size + md5) as a column in pools.
5. **Seller wallets after the gate** for BUNDLE coins (transfers past 30 min). This tells us who sells and when they pull.

## Receipts
- Scripts (container /tmp/p3): fetch2.py (logs per pool), parse.py, build.py (gate + liquidity path), sim.py (exit rules), look.py, stage.py, nft2.py (LP NFT owners), bgrid.py.
- Coins: 2,133 (BUNDLE 121, plain 1,795, other hooks 217).