# Analyst 2: 20 iterations, round 1 (9 Oct 2026, 11:10 Bangkok)
**First-principles question:** a coin has just touched $300k for the first time. What fact, known at that second, raises the chance it goes on to $3M (a runner, D-036)?
**Data:** research-box /opt/ppcopy/e1-runners.csv: every pool that peaked at $300k+ (May to 7 Oct), 6,234 with clean caps (2,231 runners = peak $3M+, rest control $300k-3M). Price path: /opt/sw/e1/e1.db sw (windows of ~33 min).
**Dumb baseline (D-032):** buy every coin at its first $300k touch. P($3M) = 27% (since 30 Sep: 21%).
**Out-of-sample check:** half A = before 30 Sep, half B = 30 Sep to 7 Oct. Scripts and raw: /root/fwa2/it20/it1.py, it2.py, it3.py (.out).
**Fill check:** entry at the cap at the END of the first window after the $300k touch (about 17 min late). That removes "jumped through $300k in one trade" artifacts.

## EARLY STAGE (stage 2): A2-L8366, launcher 0x8366a39cc670b4001a1121b8f6a443a643e40951 (v4, fee 3000 / 500 / dynamic)
- Runner rate at the $300k touch: 51% vs 27% baseline (lift 1.85; half A 1.91, half B 1.70). n=576.
- Late entry (+17 min, entry median $468k-$600k), same exits for baseline, stop 0.5, 24 h:

| exit | half A: L8366 vs baseline | half B (30 Sep-7 Oct): L8366 vs baseline |
|---|---|---|
| TP 2x | 1.12x vs 1.02x (n 303) | **1.55x vs 1.17x** (n 77, win 81%) |
| TP 3x | 1.27x vs 1.12x | **1.92x vs 1.32x** |
| TP 5x | 1.42x vs 1.24x | **2.30x vs 1.45x** |
| fee 3000 only, TP 5x | 1.52x | **2.93x** (n 50, win 80%) |

- Beats the baseline on **7 of 7 days** in half B (1.34-1.73x vs 1.09-1.30x at TP 2x).
- **Not covered by our bots:** the mc5 paper twin (/opt/sw/v3mc5.db) has 0 trades on this launcher.
- **Who loses:** buyers who arrive after the first doubling and hold. These coins die within a day: mean cap at +24 h is 0.03x of our entry. We must exit on the way up; a hold-forever rule loses.
- **Kill if:** (1) tick-level re-test (entry +60 s after the $300k touch, real swaps) does not beat the baseline; (2) outside money test: non-fleet wallets cannot sell at 2x+ in these pools; (3) forward paper under 1.2x or under the baseline over 40 trades.
- **Caveats:** windows are coarse (~33 min); TP fills assumed at 0.9 x TP; half B is only 8 days. Half A edge is small (1.12 vs 1.02).
- **Next:** tick-level re-test on v4 swaps (PP v4sender.db) + outside money test. Then the brief for the test manager.

## Side finding for Analyst 1 (mc5)
- The mc5 universe is launcher 0x73991a25c818bf1f1128deaab1492d45638de0d3 (19 of 24 twin trades mapped). It has the **lowest runner rate** of the big launchers: P($3M | $300k) 14% overall, **5% since 30 Sep** (baseline 21%). Median peak after a late entry 1.25x, 5x hit 5%.
- mc5 makes money as a quick trade, but it cannot move runner capture (D-036). Relevant to MC5GATE and the D-035 stop review.

## The 20 iterations
| # | idea (known at the $300k touch) | result | verdict |
|---|---|---|---|
| I1 | venue | v2 lift 1.49, but only 21 v2 coins since 30 Sep (v2 launches dried up) | KILLED (venue gone) |
| I2 | venue + fee | v4 dynamic-fee (8388608) "75% runners" = glitch caps (late-entry median $708M) | KILLED (artifact) |
| I3 | hooks | zero-hook v4 lift 1.13, half B 0.51 | KILLED (unstable) |
| I4 | launcher | 0x8366a39cc670: lift 1.85, stable A/B | **LEAD -> A2-L8366** |
| I5 | contract kit hash | same as launcher (kit bd3881180b54 = 0x8366) | duplicate of I4 |
| I6 | coin age at $300k | 2-24 h lift 1.39; late entry half B: entry $1.8M, peak 1.10x | KILLED |
| I7 | hour of day | 0.92-1.04 | KILLED (no signal) |
| I8 | weekday | flips sign between halves | KILLED |
| I9 | $100k -> $300k in under 30 s | lift 2.13, but late entry median $3.3M, peak 1.09x | KILLED (artifact: unbuyable jump) |
| I10 | raw touch more than 1 h before clean touch | entry $1.4M, peak 1.15x | KILLED (artifact) |
| I11 | new holders in the 1 h before $300k | only 252 coins have the field | DATA GAP |
| I12 | holders at $300k | 1k+ holders lift 1.93 on n 36 | DATA GAP (ask PP to fill holders at touch) |
| I13 | market heat (other $300k touches, prior 1 h) | 0.95-1.04 | KILLED |
| I14 | launcher momentum (its runners, prior 24 h) | 10+: lift 1.56 in A, 0.00 in B | KILLED (fails out of sample) |
| I15 | global momentum (runners anywhere, prior 6 h) | 12+: lift 1.62, but no half-B data | PARKED |
| I16 | kit already produced a runner | 0.97 vs 1.00 | KILLED |
| I17 | ladder at $1M: P($3M given $1M) | 55%, flat by age (53-60% from <30 min to >7 d) | EARLY STAGE 1: old coins at $1M are as good as new ones; needs a fill check |
| I18 | speed $300k -> $1M | <2 min: 80% runners, likely jump artifact | PARKED (needs tick data) |
| I19 | baseline by week | P($3M) fell from 41% (late Aug) to 22-24% (Sep-Oct) | NOTE: always compare within the same week |
| I20 | late bloomers ($300k touch > 24 h after birth) | 16% of all runners; P($3M) 32% vs 27% | EARLY STAGE 1: universe gap. Bots that only watch fresh launches miss 1 in 6 runners |

## Housekeeping
- research-box disk was 100% full this morning; freed 3.1G of my killed-idea files; board #52 to the Manager for the rest.
- PP R6 tape (/opt/ppcopy/r6tape.db) has landed: A2-USDG4 can start next run.