Brief
Record a 20-minute stats-round mock: one CUPED/Simpson/permutation item, one AB-test49 readout defense, and two follow-up grillings — all narrated.
Dataset
77-derived item bank (L343/q901, L345/q902, L349/q903) + your AB-test49 writeup from the bridge/P0 work.
Starter guardrails (plan 88)
Mock template (item, readout, follow-ups); recording checklist; written self-grade against the rubric within 24 hours.
100-pt rubric
Item solutions 30; readout structure (SRM/peeking/decision) 30; follow-up handling 25; time discipline 15.
Deterministic checks -> LLM review
Checks: recording present + <= 25 min; both items attempted; self-grade filed. LLM review grades readout quality and grilling composure.