# A2.2 to A2.5: results (6 Oct 2026, Bangkok)

## Read this first
- **A2.2 losers:** 38 of 59 trades lost. 28 lost about 43%, 3 lost about half, and 7 went to near zero.
  - The near-zero ones are 4 price crashes and 3 pulls, all straight through the stop.
  - With every stop filled 30% worse, the result is still 1.80x.
  - Even if every loser went to zero, it is still 1.59x.
- **A2.3 nearby cutoffs:** all 72 combinations are above 1x in both halves (lowest 1.39x, highest 2.43x). The edge is broad, not one lucky point.
- **A2.4 hold a slice with the ladder:** no stable gain. The first half gets worse, the second half better. Keep "sell 100% at 5x". The 30% bag is a paper variant only.
- **A2.5 the $20M coins:** 19 of 24 were pulled later. But the price was usually already about 99% down from its top by then. The danger is the crash, not the pull.

## A2.2 How bad are the losers
- 59 trades: 21 wins (19 at the 5x target, 2 timed out above 1x), 38 losses.
- **Losers by size:**
  - 28 at 0.5x to 0.6x: a normal stop, about −43% after fees and impact;
  - 3 at 0.1x to 0.5x;
  - 7 below 0.1x.
- **The 7 near-zero trades:**
  - 4 crashes straight through the stop in one move (0.003x to 0.12x);
  - 3 pulls (0x978177408c2f741c24ddf5ea9eb14cecbf9b3926, 0xed7bc12c9c6e6484e17f7204c16a57a4e2792668, 0xa58ae1c2b27229aade2c3017934341dc4ecc2375).
- **Stop fills made worse** (every stop sold X% below the stop price):

  | Stop fill | First half | Second half | All |
  |---|---|---|---|
  | As backtested | 1.86x | 1.90x | 1.88x |
  | 10% worse | 1.83x | 1.87x | 1.85x |
  | 20% worse | 1.80x | 1.85x | 1.82x |
  | 30% worse | 1.77x | 1.82x | 1.80x |

- **Worst case, every losing trade = 0:** 1.55x and 1.64x (1.59x all).
- **Why:** the money comes from the 5x hits (1 in 3). The losers cost at most 1x each.

## A2.3 Nearby cutoffs
- Grid (both halves each):
  - gate $750k, $1M, $1.5M;
  - age 1 h, 1.5 h, 2 h;
  - top receiver share 6%, 8%, 10%, 12%;
  - army share 0.009 and 0.02.
- **Result:** 72 of 72 cells are above 1x in both halves. Range 1.39x to 2.43x.
- **Read across the grid:**
  - $750k gate: first half 2.1x to 2.3x, second half only 1.4x to 1.5x;
  - $1M: 1.8x to 2.3x in both halves;
  - $1.5M: first half 1.4x to 1.6x, second half 2.0x to 2.4x;
  - age 1 h is best at $1M (2.3x and 2.2x, but only 16 and 26 trades).
- Full table: http://159.223.88.168/hx-c9ois/sw/sw-a2-grid.csv
- **Caveat:** army share at $1.5M was counted on wallet data that stops 10 min after the $1M gate.

## A2.4 Buy with v3, hold with the ladder
- Tested on the 59 trades (19 reached 5x):
  - sell 80%, 70% or 50% at 5x;
  - hold the rest while new buyers in the last hour are at least half of hour 1's new buyers, and price stays above a floor;
  - also exit on a pull or at 72 h.
- **The rule as written** (floor = the 5x level) ends at once. The price dips under 5x right after the first touch, so it is the same as selling 100%.
- **Floor 3x, with the buyer ladder:**

  | Sell at 5x | First half | Second half | All |
  |---|---|---|---|
  | 100% (today) | 1.85x | 1.90x | 1.875x |
  | 80% | 1.79x | 2.00x | 1.90x |
  | 70% | 1.76x | 2.05x | 1.92x |
  | 50% | 1.70x | 2.15x | 1.94x |

- **Without the buyer ladder (price only), 70%:** 1.65x and 2.06x. So the ladder helps.
- The bag lasts about 1 h on average. Best single bag: 10x the stake.
- **Verdict:** the halves disagree. Keep 100% at 5x. Run a 70/30 split as a second paper book and decide after 14 days.

## A2.5 Do the big winners get pulled
- 24 new v3 coins reached sellable $20M (13 Sep to 6 Oct).
- **19 were pulled for good** (the pool's money never came back above 30%).
- **When:** median 23.5 h after start, and 15 h after first reaching $20M.
- **Price just before the pull:** median 1% of the coin's top. They crash first (crew selling), then pull late.
- **Pulled near the top** (90%+ of the peak): 3 coins:
  - 0x2627cbc6880fcc356a49ce70aee59e18baa46224;
  - 0x60042b65a0c0a90ba75d17879850b6f99d9422eb;
  - 0xee049979960d5e42b389b941ec45790ca7b6e1f7 (HI).
- **Not pulled:** 5 coins (0x61a10e8556bed032ea176330e7f17d6a12a10000, 0xc82fb80ab69f68ca9f885f43f2b0cde0b499102c, 0x8e41fd25e85e624560fa50ff9f117a02cd30b8b3, 0x9ba4d84ac1571a0788b2bb7a6412e73529d12dc6, 0xc38c332012a9116dcade0c7f43b5d3c71799a9f8). The last three are too young to judge.
- **What it means:**
  - Holding past 5x needs a sell trigger on the crash (the price floor). A pull guard alone is not enough.
  - `w2_viral` should keep its "unpullable pools only" filter. In v3 these pools are pullable.
- Table: http://159.223.88.168/hx-c9ois/sw/sw-a2-big20.csv
- **Data note:** 4 coins show broken top prints above $500M (bad supply or one-off prints). Their pull timing still counts.

## Not done yet
- A2.6: waits for PP3/PP4 (about 7 days of fair data).
- A2.7: waits for A1.5 (crew overlap).

## First principles score
- A2.2 9/10;
- A2.3 8/10 (the army data is cut off at the $1.5M gate);
- A2.4 7/10 (19 bags, the halves disagree);
- A2.5 8/10 (my pull detection is my own, not PP's `liq` table).

Version 1, 6 Oct 2026 Bangkok. Latest: http://159.223.88.168/hx-c9ois/sw/sw-a2-results.md