# h100 hand rebuild + busy60 replay (testmgr2, 10 Oct)
## h100 hand rebuild (#778)
Independent script (rebuild.py, reads raw v4tape swaps.jsonl/pools.jsonl, does not import h100.py): entry = first swap at >= first+63 s, price from sqrtP (quote-side inverted), fee 0.01% + 3% slippage per side, pull = liq <= 50% of peak, TP 3x capped at target, stop 0.5.
Rebuilt 4 arm-A trades (3 TP, 1 quiet): entry time, entry price, exit type and net_x (2.9097 = 3 x 0.9999 x 0.97 vs csv 2.91) all match the csv. Quiet trade 0xf50228748f: no liquidity removal logged, last-price value 1.89x, counted 0 (conservative).
Arm A: 13 trades, 10 TP, 3 quiet, 0 stops/pulls: too good, so the tp fill is generous (filled at exactly 3x, no overshoot cost beyond 3%). Sellability (hook skim/block) is not covered: that is auditor2's eth_call simulation.
## busy60 30-day replay (#781)
Already exists (replay/busy60, 10 Oct 09:04): arm A 1.53x n85, B 1.14x n354, Z 0.98x n1389, P 1.36x n187, R 1.42x n354, Q 1.69x n261. Forward: A 1.66x n7, Z 0.77x n68. Replay A agrees with sim 1.60. Pulls invisible in v4tape (quiet = 0, conservative). Not re-run: no new data since.
audit_ready set true for h100, h100s, h3k, h3n, busy60, d0fd.