# Take profit and big runners: findings and needs (sell session, study only)

## Read this first
- **Nothing is live. Nothing runs.** This session only studies. It reads Price Predict data and never writes to it.
- **Why our 30 s picker never buys the big runners:**
  - It only scores coins that launch straight into a Uniswap v4 pool.
  - Pons coins launch on a Pons curve instead. About 4,100 of them launch per day.
  - When a Pons coin fills its curve, it "graduates" to its own pool. About 50 to 160 do this per day.
  - The picker sees neither the curve nor the graduated pool. Example: AGE ($63M peak) was never scored.
- **Pons graduated pools never got their money pulled:** 0 of 1,708. In our v4 picks, 97% of coins whose money the creator controls got pulled.
- **So on Pons, letting a part run pays.** Pons coins picked at graduation (top 10%, 114 coins):
  - sell all at 10x: +13% per trade;
  - sell half at 10x, trail the rest by 50%, stop at minus 50%: **+62% per trade**. Positive in both halves (+71% and +48%).
  - Small sample. Paper first.
- **v4 coins (our picker today): keep selling everything at 10x.** The creator owns the pool money and pulls it in minutes.
- **Biggest to-do for Price Predict:** add Pons (curve plus graduated pool) as a second launch venue, with wallet features. Full list at the end.

## Answer on the other session's $250k study
- **Agree for v4 plain coins.** The creator owns the LP NFT, and 95 to 99% are pulled within minutes. Sell at the gate, fast. Two separate checks found this.
- **Agree that "locked" alone is no reason to hold on v4.** We both saw locked v4 coins fade.
- **Bundle hold:** fine as a paper test only. Its edge shrank in the newer half, and bundle pools still get pulled within hours.
- **What it missed: Pons.** Pons graduated pools are a different lane, with 0% pulls. At $250k on Pons:
  - selling right away keeps 0.77 of a $3,000 bag (our own sell and the fee eat the rest);
  - holding with a trail keeps 0.81 to 0.82, the same in both halves;
  - so holding costs nothing, and 11% go on to a sellable $2M.

## Findings, in order of how sure we are

### A. Who can take the pool money decides how to sell (proven)
- v4 coins (3,895 checked):
  - creator wallet owns the LP NFT: 97% pulled (96% early half, 97% late half);
  - a contract owns it: 8% pulled;
  - burned: 0% pulled.
- Pulled coins: 60 of 60 had the creator wallet as owner at buy time. Kept coins: 74 of 80 had a contract or burn as owner.
- Half the pulls happen within 2% of the top, a median 3 s after the last trade. A chart cannot warn you.
- Pons graduated pools: 0 of 1,708 pulled, and liquidity never fell below half.

### B. v4 lane, sell rules (proven on 2,776 picks)
- Sell 100% at 10x.
- Also sell everything if nobody trades for 60 s, and at 30 min. In past data: +$198 per $200 trade, against +$128 for the live rule. Better in both halves.
- Any bag after 10x loses: all at 10x +7.87 per trade on 10x coins; keeping a 25% bag +6.45; holding 7 days minus 0.57.
- After 10x, only 15% reach 30x before the pull.
- The paper books value pulled coins at their last price: 12 fake wins found. The picker session should fix this.

### C. Pons lane (new, 5 study rounds)
1. **Size.** 90,810 Pons launches from 13 Sep to 5 Oct, 1.26% graduated. Median 9 min from launch to graduation, and 25% graduate within 52 s. Graduations per day fell from about 160 (mid Sep) to about 35 (early Oct).
2. **Blind buy of every graduation (30 s after it) loses.** All at 10x: minus 21% per trade. Hold 7 days: minus 60%. Yet 8.3% do 10x, 2.8% reach a sellable $2M and 0.6% reach $5M. The big ones are there, but a filter is needed.
3. **Signs at graduation +30 s are weak:** cap at graduation, fast graduation and low sell share, each with AUC 0.56 to 0.62. Signs from the runner study (holders spread out, cap at 1 to 8 h, X authors) should be added on top.
4. **At $250k:** hold and sell are equal (0.77 to 0.82 of a $3,000 bag), with a long right tail: 22% reach $1M and 11% reach $2M.
5. **Picker tests:**
   - At graduation (top 10%): the runner bag rule makes +62% per trade (see Read this first).
   - On the curve at 30 s, using trade counts only: top 1% get 7.5% full path 10x, against 0.8% base. Our v4 picker gets about 30%. The curve picker needs wallet features (alumni wallets, unique buyers, bundles, creator history), like the v4 model.

### D. Runner tracker (not trusted for holding)
- Most coins in it were found 11 to 30 days after they crossed $500k, so it holds only survivors. That is why holding looked great there (2.1x in 7 days).
- Use it only for patterns, never for hold money.

## The framework (v3, study only)
1. **At the buy, set the lane:**
   - **PULL lane:** v4, creator wallet owns the LP NFT. Also "other" adders until proven safe.
   - **HOLD lane:** Pons graduated pool, LP burned, or a known locker contract.
2. **PULL lane:** sell 100% at 10x. Sell all on 60 s with no trade, and at 30 min. Never keep a bag.
3. **HOLD lane:**
   - stop at minus 50%;
   - at 10x, sell half;
   - trail the rest by 50% from its top, up to 7 days.
   - Add runner checks for the bag later (1 h, 4 h, 8 h: cap at its high, holders rising, low churn, X authors). This waits for unbiased data.
4. **Phase 2 gate at $250k:**
   - PULL lane: sell within 60 s.
   - HOLD lane: keep a bag with the trail. It costs nothing and keeps the 11% chance at $2M.

## To-do for Price Predict (they build, I read)
| # | What | Why | Priority |
|---|---|---|---|
| 1 | Score Pons launches: curve trades from second 0 (buy and sell events from the Pons factory 0xe33e9e479df8802cb0866d5d05258bec4cf62948), with the same wallet features as aipick5 | The big runners are here; the picker cannot see them now | Urgent |
| 2 | Score every Pons graduation (pool hook 0xe5e702641ea86f4ae6cc3cdaed2b886f976be044, migrator 0x7ed598bcef8bd9edd8c97a195c6d13f40801ec7e) as its own launch moment | 50 to 160 a day, never pulled, +62% per trade with a runner bag in the test | Urgent |
| 3 | Finish chainswaps (all venues, 15 Sep to now). It was 5 to 13% done on 5 Oct | Unbiased runner paths on every venue, so the hold rules can be proven | High |
| 4 | Lane column at buy time: owner of the LP NFT (PositionManager 0x58daec3116aae6d93017baaea7749052e8a04fa7 Transfer logs for the tokenId), plus Pons, v2 or v3 lock status | One flag decides sell fast or let it run | High |
| 5 | Liquidity table per pool: every add and remove, the share removed, and who did it | Honest exits, and fixes the fake 24 h wins | High |
| 6 | Paper books: check liquidity before booking any exit; a pulled pool pays about 0 | 12 fake wins found | High |
| 7 | Pons curve tape back to July (ponsbuy, ponssell, graduation) | Only 10 days are pulled so far (3.2M trades, scratch copy in my container) | Medium |
| 8 | Hourly holders and X authors for every coin at $250k+ (Pons included) | Runner checks for the bag | Medium |
| 9 | Supply for every Pons token: 559 of 1,708 graduated tokens had none in tokens.db | Cap lines need supply | Medium |
| 10 | Names of the 3 locker contracts: 0xeff166aaf189323c58dc27ed1206eb2c37faacdf, 0x267444d099b10fb5ed7c3cc7b7c767adca574952, 0x35c5c1037eef9eb07b2826d537e4fdae61878852 | Which lockers are really safe | Low |

## What I need to finish the study
- Items 1 to 3 above. With them I can rerun every Pons test on 3x to 5x more coins and prove or kill the runner bag rule.
- About 4 weeks of forward Pons graduations, for a paper check of the HOLD lane rule.
- No money, no keys, nothing live.

## Files
- Research scratch (container): /root/sell (v4 study), /root/pons (Pons: launches, graduations, 10 days of curve trades, graduated pool paths with liquidity events).
- Droplet /opt/sellsys holds study scripts only. Nothing runs.
- Public data copies: http://159.223.88.168/hx-c9ois/sellx/

Version 3, 5 Oct 2026, about 18:45 BKK. Latest version: claude/sell-system-v1.md in the Pro Trading project. Public copy: http://159.223.88.168/hx-c9ois/sellx/sell-system-v1.md