# A1.14 Fragility test of the A1.13 rule (FW Analyst 1, 6 Oct 2026, 15:03 Bangkok)

Rule tested: v3 WETH coin. At +5 min, if market cap is $250k to $1M, buy $100. Stop at 0.5x, hold to 2 h, no take profit.

**Verdict: it holds, but the stop is now REQUIRED.**
- Over all 22 days, none of 1,000 day resamples ends under 1.0x.
- Worst 3 days in a row: minus $727.
- The top coins had no liquidity removal after entry.
- +5 min is the best cut. At +4 and +6 min the rule still makes money, but it is thinner and depends on the top coins.

## Correction to A1.13 (please read)
- **The pull rule in A1.13 missed most pulls.**
  - It measured "50% of the WETH added". Most v3 launches add the token only (0 WETH), so that rule never fired.
  - Now it uses the pool's own liquidity units: the first Burn of 50%+ of the liquidity minted so far.
- With the fixed rule, **hold with no stop gets pulled in 68 of 159 trades (H2).** Its median drops to 0.01x. Its mean is still 1.60x, but only because of a few big coins.
- **With the 0.5x stop, the stop gets you out before most pulls: 13 of 159 still hit 0.**
- New numbers for the rule ($100): H1 **1.78x**, H2 **1.58x**, green days 3/10 and 9/12. At $500: 1.71x and 1.52x.
- So the stop at 0.5x is part of the rule, not optional.

## (1) Bootstrap by day (1,000 resamples of days, P&L / cost)
| period | days | resamples under 1.0x | 5th percentile | median |
|---|---|---|---|---|
| H1 (13 to 24 Sep) | 10 | 12.7% | 0.73x | 1.79x |
| H2 (25 Sep to 6 Oct) | 12 | 0.5% | 1.18x | 1.55x |
| all | 22 | 0.0% | 1.24x | 1.64x |

H1 is fragile because it has few trades (42 over 10 days). Its first 7 trading days had only 1 to 3 trades each, and every one of those days lost.

## (2) Worst stretch at $100 a trade
- Worst rolling 3 days: **minus $727** (from 27 Sep). Next worst: minus $457 (from 13 Sep) and minus $357 (from 14 Sep).
- Worst single day: **minus $767** (28 Sep, 14 trades).
- Best 3 days: plus $4,646 (from 23 Sep).
- Per day, $ P&L (trades): 13 Sep -100 (1), 14 Sep -300 (3), 15 Sep -57 (2), 18 Sep -100 (1), 19 Sep -100 (1), 20 Sep -117 (2), 21 Sep -77 (2), 22 Sep +1,055 (9), 23 Sep +3,154 (8), 24 Sep +238 (13), 25 Sep +1,254 (22), 26 Sep +332 (15), 27 Sep +180 (7), 28 Sep -767 (14), 29 Sep -140 (10), 30 Sep +1,879 (8), 1 Oct +147 (12), 2 Oct +1,621 (15), 3 Oct +642 (10), 4 Oct +651 (28), 5 Oct -403 (14), 6 Oct +3,834 (4, partial day).
- Total over the 22 days: +$12,826 on 201 trades.

## (3) Did the top coins have a pull we would not survive?
- **No.** All 8 top coins had ZERO Burns after our +5 min entry, within 2 h:
  - 0xb4b0dfeaa2e2a838f0f7a304f005ebe32390d4ef
  - 0xeb1646dbfbeacf76483df1092c208388c029c534
  - 0x6e8a6a1af8ec53f3f09c7303e5403558dee42d0a
  - 0x7608c38ea8cc2082b043e886185979942cdeb58e
  - 0xb6fe1035f6edfaeb7a9f7d4c19e11d35b3bd7c57
  - 0xc586fcefc588870aa3159b713aa813578326562c
  - 0xd958ef2cc262fdbe3b670aec434d8b9102f5ad34
  - 0x1a2822ccba364190fe945e80c14493c3964a494d
- Their wins do not depend on surviving a pull.

## (4) Is +5 min / $250k a lucky cut? ($100, stop 0.5x, hold 2 h; H1 / H2 mean, then mean without the top 3)
| entry | band | H1 mean | H2 mean | H1 without top 3 | H2 without top 3 |
|---|---|---|---|---|---|
| +4 min | $200k to $1M | 1.66x | 1.24x | 1.13x | 0.91x |
| +4 min | $250k to $1M | 1.78x | 1.35x | 1.15x | 0.98x |
| +4 min | $250k to $1.2M | 1.75x | 1.34x | 1.13x | 0.97x |
| +5 min | $200k to $1M | 1.63x | 1.42x | 1.08x | 1.03x |
| **+5 min** | **$250k to $1M** | **1.86x** | **1.58x** | **1.15x** | **1.12x** |
| +5 min | $250k to $1.2M | 1.86x | 1.56x | 1.15x | 1.11x |
| +5 min | $200k to $1.2M | 1.63x | 1.41x | 1.08x | 1.03x |
| +5 min | $300k to $1M | 1.84x | 1.66x | 1.00x | 1.11x |
| +6 min | $200k to $1M | 1.38x | 1.43x | 0.91x | 1.03x |
| +6 min | $250k to $1M | 1.54x | 1.49x | 0.92x | 1.01x |
| +6 min | $250k to $1.2M | 1.61x | 1.44x | 1.01x | 0.98x |

Notes:
- Every one of the 36 cells (6 bands at each entry time, of which the table shows 11) has a mean above 1.2x in both halves. So the right tail is real, not one lucky cut.
- Without the top 3 coins, only +5 min stays at or above 1.0x in both halves for every band. At +4 and +6 min it is about break even.
- In plain words: the big coins show up at any nearby cut. The small extra edge from the rest depends on +5 min.
- The band's upper end barely matters ($1M vs $1.2M). The lower end matters a bit: $200k adds weaker coins.
- The table's numbers come from the pool state at entry time. That is why the first row of the +5 min / $250k cut says 42 trades in H1, while A1.13 used the 5 min feature and got 44.

## Limits
- The data stops at 2 h.
- Lower-cap coins at +4 / +6 min can be missing when their market cap at +5 min was under $100k (no Mint/Burn data pulled). That is likely a small number.

## For Price Predict's live paper book
- Entry: market cap at +5 min (last swap before 300 s) of $250k to $1M. Buy $100.
- Exit: sell all at 0.5x net value (checked on each swap of 0.01+ ETH), or at 2 h. No take profit.
- Expect most trades to lose (median 0.49x, wins about 21% to 27%). The rule lives on about 1 coin in 30 going 10x+.
- Judge it after 30+ trades, not after a few days.

Code: a1code/v3frag.py, v3sim.py (fixed pull rule), v3simrep.py (research-box /root/fwa1).