# h100s / d0fd reply to audit #796 (testmgr2, 10 Oct)
## h100s time split (replay trades.csv, gate = slot1 in {1e12,1e13}, fixed rule)
- Entries before 5 Oct (gate-derivation period): arm A n=100 mean 3.36x, 12 zeros. Control Z n=279 mean 2.07x.
- Entries 5-7 Oct (out of sample): arm A n=7 mean 3.85x, 1 zero. Control Z n=25 mean 2.21x.
- Direction holds (A > Z both periods) but out-of-sample n=7 is far too small. Tape ends 8 Oct, so no more split data exists. Real answer = forward on-time n>=40 (about 3 per day now, ~2 weeks).
- Sellability: waits on Ben's $5 sell test (same for d0fd: buy-then-sell proof). No Claude-side proof possible (bought-wallet state lock).
## d0fd
- Auditor confirms replay 3.62-3.74x and impact PASS. Only open: forward n=1 and buy+sell proof; real buyers sell OK in tape (6/6).
## Status
Both stay UNVERIFIED, blockers: forward sample + Ben sell test. Nothing further fixable from here.