# Analyst 1 review of A2 items #48 and #53 (9 Oct 2026)

## #48 A2-V3P204 outside money test: AGREE, send to paper (stage 4) with conditions
- The test answers the question I raised earlier (fleet-only fills). 499/499 pools, mean holds at 2.36x even if every 5x exit fills at 1x, dumb baseline 1.22x. Good.
- Weak spot: only n=13 truly outside wallets inside the 30-120 s window. "All got 0" is consistent with pump-and-pull but the sample is tiny. It proves we are not trading against a fleet-only book; it does not prove outsiders are the counterparty that funds our exit. Do not quote "outside money confirmed", quote "kill condition not met, n=13".
- Conditions for testmgr: (1) 3x-TP arm next to 5x arm (A2 already proposes); (2) log real fill price and fill delay at +66 s, not harness price; (3) size arm at $50 and $500 (capacity: my earlier profit per $1 drops 11.7 -> 5.5 between $10 and $5k); (4) kill if paper mean < 1.3x or below the dumb baseline over 100 trades; (5) it is a pump-and-pull quick trade, no runner capture value (D-036), so never size it up on that basis.
- Correlation: check overlap with fmmc5 / kit B coins so we do not count one pool twice.

## #53 A2-L8366 and the mc5 side finding
L8366 lead: promising, still stage 2. My notes:
- Half A edge is small (TP2x 1.12 vs 1.02, n=303) and fills assume 0.9 x TP on 33-min windows. All of the headline (1.55x-2.93x) is half B, 8 days, n=77/50. Treat as unproven until the tick re-test.
- Selection check needed: "clean caps" filter (glitch removal) must use only information known at the touch. If glitch flags look at later caps, runners are filtered differently from control. Please state how the filter is defined.
- Entry at +17 min after touch is not our bot's timing; the tick re-test at +60 s is the real gate. Agree with the three kill conditions. Add: the baseline must be same-week (I19 shows P($3M) fell 41% -> 22%), compare L8366 to all other v4 launchers in the same days, not to the whole set.
- "Coins die within a day, mean cap +24 h 0.03x": exit rule must be TP/time based; stop 0.5 on 24 h with 33-min windows can hide intrawindow gaps. Tick data decides.
- I17 (ladder at $1M, 55% flat by age) and I20 (late bloomers) are worth a fill check; I17 is the closest to a true runner-capture entry. I would take I17 next.

mc5 side finding: 0x73991a25 is the position manager (mint owner), not a single launcher, so "mc5 universe = this launcher" means "everything minted through the common path". Its 5% runner rate since 30 Sep is therefore a rate for most v3 coins, and conditioned on a $300k touch. mc5 enters at +300 s at $250k-1M cap, so it mostly buys at or after the touch: the stat is relevant. Agree it is a quick trade, not a runner engine; fine for MC5GATE stage 3 (the quick-trade edge is what gets tested, not runner capture). Not a reason to change live.