Pre-FOMC drift on SPY self-test, 2026-11-04T11:50:29.091Z

1. The announcement list
  [ok] eight scheduled announcements a year 1994-2026, seven in 2020
  [ok] first announcement 4 Feb 1994; September 2003 on the 16th (the 15th, an agenda entry, is excluded)
  [ok] meetings across two months (Jan/Feb and Oct/Nov 2017, Jul/Aug 2018) end on their second day
  [ok] cancelled, unscheduled and conference-call days are not in the list (March 2020, 3 Jan 2001, January and October 2008)
  [ok] the post-election Thursday meetings (8 Nov 2018, 5 Nov 2020) are in; the CSV holds the same list

2. Hand-built days
  [ok] one trade, close of 7 Nov to close of 8 Nov 2018, dividend on the day: gross 1.2871%
  [ok] overnight part (dividend at the open) and day part; weekday Thursday
  [ok] reference identical
  [ok] a Tuesday announcement (16 Dec 2008): bought at the close of Monday 15 Dec
  [ok] a file starting on the announcement day cannot trade it
  [ok] Day before: 1 trade of 1 day(s), 2018-11-06 -> 2018-11-07
  [ok] Two days: 1 trade of 2 day(s), 2018-11-06 -> 2018-11-08
  [ok] Day after: 1 trade of 1 day(s), 2018-11-08 -> 2018-11-09

3. Control pools and draws
  [ok] same month and weekday, never an announcement day: the pool for 8 Nov is 2018-11-15
  [ok] drawn day 15 Nov: bought at the close of 14 Nov
  [ok] no other November Thursday: the fallback pool is the same weekday in the quarter (25 Oct)
  [ok] drawing the real day gives back the real trade

4. Real data
  [ok] PF: 261 trades, reference 261, identical
  [ok] every scheduled announcement in the file's range is traded (261)
  [ok] every pool is non-empty, same weekday and month, without announcement days (fallbacks: 0)
  [ok] drawing the real days gives back the 261 real trades

COUNTS ONLY: 261 trades (DEV 96, VAL 80, HOLDOUT 85); by weekday of the announcement Tue 85, Wed 166, Thu 9, Fri 1

PASSED