# Second window plan v2 (6 Oct 2026, 12:40 Bangkok)

## READ THIS FIRST (for Ben, then for Price Predict)
- From: the second window supervisor (review session). Second window = any buy later than the first seconds after launch.
- v2 adds 5 more research rounds on the big winners (coins that went to $1M, $5M, $20M). Data: every Pons graduation from 5 Aug to 3 Oct 2026 (9,986 coins: 505 reached $1M, 97 reached $5M, 20 reached $20M), with hourly prices for 2,071 of them.
- **Rule 1 (unchanged): only buy later in a pool the creator CANNOT pull.** In creator-held pools, 95% to 99% are pulled minutes after $250k.
- **The big winner pattern is about the LANE, not the coin.** A lane is a launch site plus the coin it is paired with: Pons with ETH, Pons with NVDA, Pons with USDG, and so on.
  - Coin traits (name, launch hour, weekday, curve speed, buyers, X) barely move the odds. Earlier sessions found the same.
  - Lane age and lane rank move them 5x to 30x.
- **Nugget 1: NEW LANES make the big winners.** Share of Pons graduations that reached $5M, by week:

  | Week | Reached $5M |
  |---|---|
  | First week of Pons (5 to 9 Aug) | 8% |
  | Late August | about 1.5% |
  | Early September | 1.1% |
  | Late September | 0.25% |

  - 16 of the 20 $20M coins graduated before 7 Sep.
  - Inside new pairings (for example the first days of a new stock quote), the first 3 days give 3.7% reaching $5M, against 0.35% to 0.7% for the same pairings when older (after 7 Sep). Before 7 Sep: 2.3% against about 1.1%.
- **Nugget 2: LANE LEADER. Buy the FIRST coin in each lane to close an hour at $1M or more.**
  - 31 cases (about 0.7 a day). Average result **+71% per trade** at $1k, after fees and real pool slippage. Exits: sell half at 2x, trail the rest 50% from its top, stop at −50%, 14 days max.
  - Positive in both periods: before 7 Sep +91% (20 trades), after +35% (11 trades, median +47%).
  - Taking out any single coin still leaves +25% to +77%. Luck range (90%): +1% to +159%.
  - It works step by step: 1st coin in its lane +71%, first 2 +43%, first 3 +25%, first 5 +17%. All coins crossing $1M together lost money after 7 Sep (−24%).
  - Size: these pools hold about $50k at $1M, so $1k to $5k per trade fits (+44% at $5k before 7 Sep, +5% after).
  - Why: each lane ends up with one "main" meme (the same thing the long.xyz lane map showed: one meme per stock). The first coin to $1M usually takes that seat.
- **Nugget 3 (from v1, small): Pons graduated coins at $100k with a low sell share:** +18% per $100 trade on unseen days.
- **What did NOT work** (killed, both halves): buying every coin that crosses $250k / $500k / $1M (loses after 7 Sep: −24% to −34%); "no other coin crossed in the last 24 h"; fast vs slow climbers; curve speed; launch hour or weekday; ticker length; quote type alone (ETH 5.9% to $1M, stock 4.7%, USDG 3.3%).
- All times Bangkok. Never shorten addresses. Paper only until Ben says go.

## Part A: what Price Predict watches today (6 Oct, 10:40 Bangkok)
| What | Status |
|---|---|
| Pons curve, graduations (ponstape) | LIVE |
| Paper buys at 30 s on Pons, v2, v3, repools (venuepick) | LIVE (blind books lose: Pons curve −20%, v2 −15%, v3 −33%) |
| All-venue swap history (chain/b*.db) | **STALLED** since 5 Oct 20:50 (liqbuild2 running 15.5 h) |
| Liquidity table `liq` | **NOT BUILT** |
| After-launch watcher on all venues | **MISSING** |
| Lane table (lane age, lane rank, first coin to $1M per lane) | **MISSING** |

## Part B: the plan (Price Predict builds, the review session supervises)
1. **Unblock the all-venue swap history today.** Run chain/b*.db beside liqbuild2 at low priority, or finish liqbuild2 first. Done = 15 Sep to now covered, and the `liq` table exists.
2. **Hold-lane flag on every pool today:** `holdlane` = 1 if the pool cannot be pulled. That means:
   - Pons graduated hook 0xe5e702641ea86f4ae6cc3cdaed2b886f976be044;
   - LP NFT sent to 0x000000000000000000000000000000000000dead, or burned;
   - LP held by locker 0xeff166aaf189323c58dc27ed1206eb2c37faacdf, 0x267444d099b10fb5ed7c3cc7b7c767adca574952 or 0x35c5c1037eef9eb07b2826d537e4fdae61878852;
   - launchpad hook adders that never pull.

   No second window buys anywhere else.
3. **Lane table, live** (new, top priority after step 1).
   - Lane = venue (Pons graduated, long.xyz hook 0x4e3468951d49f2eea976ed0d6e75ffcb44a9a544, bankr, v3 with burned LP, other launchpads by hook or factory) × quote token.
   - Per lane: first launch time (lane age), launches per day, graduations per day, and every coin's first hourly close at $250k, $500k, $1M and $2M, with its rank in the lane.
   - A cross counts only if $20k or more was bought at or above that level (a money test, not a price print).
   - Publish it as one public file.
4. **Paper book `w2_leader`** (live as soon as step 3 runs):
   - Entry: the 1st coin in each hold-lane lane to close an hour at $1M or more (money tested). Buy $1,000 at that close.
   - Exit: half at 2x, trail the rest 50% from its top, stop at −50%, max 14 days.
   - Price every fill with the pool's depth and 1% fee.
   - Log ranks 2 and 3 as control books (`w2_leader2`, `w2_leader3`).
5. **New lane alarm** (live with step 3). Flag any lane less than 3 days old that already has 3+ graduations, or a coin at $250k.
   - Paper book `w2_newlane`: $100 on every hold-lane coin in a flagged lane that reaches $100k, for the lane's first 7 days.
   - Exits as in step 4.
6. **Keep from v1:**
   - Paper book `w2_pons100`: Pons graduated, first swap at $100k, sell share since graduation 30% or less, $100. Exits as in step 4.
   - Climb table for hold-lane coins from $100k: every 5 min for 24 h, then hourly for 7 days. Cap, pool depth, buy and sell counts and ETH, distinct buyers (tx.from), new wallets, holders, top 10 share, X authors.
7. **Backtest the lane leader rule on every venue** once step 1 lands: long.xyz stock lanes, bankr, v3 with burned LP, from July if the backfill allows. Report per venue. This is where the 31-case sample becomes 100+.
8. **Go or kill around 20 Oct:**
   - Go: `w2_leader` positive after costs and without its best coin, and the backtest from step 7 agrees.
   - Then Ben places the first real trades: $500 to $1,000 per lane leader.

## Part C: do not
- No second window buy in a pool the creator can pull.
- Don't buy every coin that crosses $1M: as a group they lost money after 7 Sep (−24%). Only the first one or two in a lane.
- Don't size above what the pool depth allows (about 2% of the pool's ETH).
- Don't trust survivor lists for hold results.
- Don't trade live until Ben says go.

## Part D: receipts
- Data: http://159.223.88.168/hx-c9ois/fw/univ_pons.json.gz (9,986 Pons graduations with peak caps, plus hourly prices for 2,071) and http://159.223.88.168/hx-c9ois/sellx/pons_*.jsonl.gz (every swap of 1,708 graduated pools, 13 Sep to 4 Oct).
- Money tests:
  - hourly closes;
  - 1% fee each way;
  - slippage from pool depth (4.4 ETH at a $56k graduation, growing with the square root of the cap);
  - exits as in step 4.
- Scripts (review session container): /root/w2/g1.py to g4.py, it1.py to it5.py.
- Caveats:
  - The peak caps are price prints. The money results use hourly closes, so they are lower than the peaks.
  - 31 lane leader cases is a small sample. One coin (NOVAAI, +1,440%) moves the average.
  - Pons only so far.

## Part E: send back to Ben
- Steps 1 to 3 done (yes or no, Bangkok time).
- The lane table's public link.
- The daily line of `w2_leader`, `w2_newlane` and `w2_pons100`: trades, money result, and the result without the best coin.

Version: v2, 6 Oct 2026 12:40 Bangkok (v1 11:30: hold lane rule and the $100k sell share rule). Latest version: http://159.223.88.168/hx-c9ois/brx/second-window-plan.md