# A1.21 mc5 by launch kit: split, exits, entry (FW Analyst 1, 6 Oct 2026, 16:42 Bangkok)

**Verdict: the kit split is real information, but it does NOT improve mc5.**
- Holding kit coins longer (a 50% trail armed at 3x or 5x, up to 48 h) **loses to the plain 4 h stop** on mc5 entries. Kit A H2: 1.53x to 1.55x vs 2.03x. Kit B H2: 0.81x to 0.82x vs 1.12x.
- Skipping non-kit coins, or a wider $150k to $2M band for Kit B, both **cut $ a day.**
- The H1 pick ("trail from 3x on kits") rested on only 2 H1 kit trades. It failed on H2: $558 vs $979 a day, and 3.8% of bootstrap resamples under 1.0x vs 1.0%.
- **Base stays: buy at +5 min, $250k to $1M, dust guard, stop 0.5x, sell at 4 h. Every kit.**
- Why this differs from Virality V8: V8 buys at the $1M gate and decides at 5x. mc5 buys earlier, before the gate, at a fraction of the price. Most of a kit coin's run is already inside our 4 h, and after 4 h kit coins fade or get dumped like the rest. Holding 48 h: Kit A 0.32x, Kit B 0.42x.

## Labels
- Method (A1.14): keccak of the bytecode of the pool-create tx's "to". Kit A = 0e81be52d5cf, Kit B = 89885b75fd29, PM = to 0x73991a25c818bf1f1128deaab1492d45638de0d3.
- 818 pools with cap $100k+ at +5 min: PM 533, other 96, Kit B 95, Kit A 94.
- Labels: research-box /root/fwa1/v3/kit.json.
- Swaps plus Mint/Burn extended to 48 h for kit pools in $150k to $2M.

## (1) Split by kit (stop 0.5x, 4 h, $250k to $1M, dust guard, $100)
All coins with 4.5 h+ of data (6 Oct included):

| kit | H1 n / mean / $ a day | H2 n / mean / median / $ a day |
|---|---|---|
| Kit A | 2 / 0.51x / -$9 | 62 / 2.03x / 0.48x / **$532** |
| Kit B | 0 | 63 / **2.21x** / 0.49x / **$635** |
| PM | 15 / **5.97x** / $678 | 5 / 0.18x / -$34 |
| other | 14 / 1.23x / $29 | 17 / 3.00x / $283 |
| all | 31 / 3.48x / $698 | 147 / 2.16x / $1,416 |

- H1 money came from position-manager launches. They disappeared in H2 (5 trades, 0.18x).
- H2 money is Kit A (24 Sep to 1 Oct) plus Kit B (from 1 Oct), as A1.14 found for the gate.
- **Kit B through 4 Oct only (48 h set) was 1.12x (n35).** Its money sits mainly in 5 and 6 Oct (for example token 0xc38c332012a9116dcade0c7f43b5d3c71799a9f8 on 6 Oct).
- If Kit B stops, mc5 loses most of today's flow. The health alarm (fw/mc5-health.txt) will show that.

## (2) Exit by kit (48 h set: first swap before about 4 Oct 17:00 Bangkok so every exit can finish; H2 = 25 Sep to 4 Oct)
| coins | exit | H2 n | mean | $ a day |
|---|---|---|---|---|
| Kit A | **4 h (base)** | 62 | **2.03x** | **$638** |
| Kit A | 50% trail armed at 3x, max 48 h | 62 | 1.53x | $327 |
| Kit A | 50% trail armed at 5x, max 48 h | 62 | 1.55x | $343 |
| Kit A | hold 48 h | 62 | 0.32x | -$424 |
| Kit B | **4 h (base)** | 35 | **1.12x** | **$42** |
| Kit B | trail from 3x / 5x, max 48 h | 35 | 0.81x / 0.82x | -$68 / -$64 |
| Kit B | hold 48 h | 35 | 0.42x | -$202 |
| non-kit | **4 h (base)** | 22 | **2.36x** | **$299** |
| non-kit | sell all at 5x, else 4 h | 22 | 1.33x | $73 |
| non-kit | trail from 3x, 48 h | 22 | 2.40x | $309 |

- H1 has 2 kit trades, so the H1 side of the kit exits is empty.
- Non-kit H1: 4 h 3.68x vs 5x target 2.13x. Selling at 5x loses in both halves.

## (3) Combined books (48 h set; $ a day; worst 3 days)
| book | H1 | H2 |
|---|---|---|
| **base (all 4 h)** | $698, -$357 | **$979, +$700** |
| base + Kit B wide $150k to $2M | $698, -$357 | $913, +$700 |
| skip non-kit | -$9, -$98 | $680, +$466 |
| kits trail from 3x, non-kit 4 h (H1 pick) | $707, -$357 | **$558, +$132** |
| kits trail from 5x, non-kit 4 h | $698, -$357 | $578, -$170 |
| kits 4 h, non-kit sell at 5x | $288, -$357 | $753, +$700 |

- Bootstrap H2 (1,000 day resamples): base 1.0% under 1.0x (5th pct 1.22x); H1 pick 3.8% (5th pct 1.03x).

## What this means
- **No change to the mc5 rule.** Keep kit labels for tracking only. Track a daily Kit B count beside the health alarm, because Kit B is now the flow.
- Kit holds are a 2nd window (gate entry) idea. They do not carry over to a +5 min entry.

Code: a1code/v3kit.py, v3kitsim.py, v3kitsplit.py (research-box /root/fwa1).