# Sample construction filters (CLAUDE.md §6). All liquidity screens use LAGGED data only.
baseline:
  session: regular_hours_only          # 09:30-16:00 ET, calendar-truncated half days
  min_price_lagged: 5.0                # USD, prior-month median close
  min_bars_per_day_lagged: 100         # prior-month median RTH bars/day
  min_history_days: 60
  exclude_suspected_split_window_days: 1   # +/- days around detected split jumps
  duplicate_rule: aggregate            # volume-sum, max(high), min(low), first open, last close
universes:
  all_instruments: {}
  liquid_top500:  { rank_by: lagged_dollar_volume, top_n: 500 }
  liquid_top1000: { rank_by: lagged_dollar_volume, top_n: 1000 }
  liquid_top3000: { rank_by: lagged_dollar_volume, top_n: 3000 }
periods:
  pre_decimalization: ["1992-01", "2001-01"]
  post_decimal_pre_regnms: ["2001-02", "2007-02"]
  regnms_to_2019: ["2007-03", "2019-12"]
  modern: ["2020-01", "2026-03"]
  # NOTE: period cuts are provisional; main analyses use flexible year effects (CLAUDE.md §13)