# h3n arm A 5x plateau: explained (testmgr2, 10 Oct)
Not an operator price level. It is my reserve-impact haircut.
- Net = gross x C^2 (0.8836... = 0.94) x imp, imp = (1 - 0.04/(R+0.04))^2, R = prior net buys, floored 0.01.
- The 80 rule rows are almost all R = 0.10 (the rule threshold, so R sits at the floor of the filter). imp = (0.1/0.14)^2 = 0.51 for all of them.
- Gross price at +595 s is ~10.3x median (78 time exits), x 0.94 x 0.51 = ~5.0x. Different pools, same gross ~10x (same-template launches: same cap_buy 22,334 USD, 4 pre-swaps) and same R, so same net.
- Example pool 0x41be4c90..: price path from tape: 1.0x at entry, 10.5x at +595 s, built by 0.1-0.4 ETH buys, with sells in between (real two-way flow, not a fixed-price operator).
- Depth check (liq x sqrtP from the tape, virtual ETH reserve at exit): median 26.7 ETH, min 9.4 ETH. A 0.04 ETH sell has <0.5% impact. With this reserve the honest net mean is 8.7x, so the 4.5x is conservative on depth, not a mirage.
- What it does NOT prove: sellability (honeypot/hook blocking sells) and that outsiders sold at ~10x. Pool sells exist in the tape (negative amount0 up to 0.7 ETH at the exit window), which supports it. Still open: Ben $5 sell test, hand rebuild of 3 trades.