Gold night/day self-test, 2026-10-21T10:33:44.521Z

1. Hand-worked series
A. Mon 6 to Wed 8 Jan 2020: 08:20 and 13:30 each day (Mon 100/101, Tue 103/102, Wed 104/104), plus 10:00 and 16:00 minutes that must be ignored
  [ok] day shorts: -1.3000, 0.6796, -0.2885
  [ok] night longs (Mon->Tue, Tue->Wed; Wed has no exit day): 1.6832, 1.6667
  [ok] GN-LS: 5 trades alternating, flips on the same minute
  [ok] financing: one rollover on a night trade (A and B), none on a day trade
  [ok] reference identical (3 variants)
B. weekend: Fri 10 Jan 13:30 (200) -> Sun 12 Jan 18:00 (210, a "13:30 or later" minute that must not open a second position) -> Mon 13 Jan 08:20 (205)
  [ok] one night trade Fri -> Mon, 200 -> 205, one rollover (Friday 17:00)
  [ok] reference identical
C. holiday: Thu 16 Jan 13:30 (300); Fri 17 Jan has only an 18:00 minute; Mon 20 Jan 08:20 (306); a missing 08:20 minute on Tue 21 Jan (first minute 09:00 at 310)
  [ok] nights: Thu -> Mon (Friday skipped), Mon -> Tue at the 09:00 minute
  [ok] Tuesday day trade from the 09:00 minute to the 13:31 minute
  [ok] reference identical (3 variants)
D. control: offset 0 returns the real trades; offset 60 enters one hour later (here 14:30) and holds the same number of minutes
  [ok] offset 0 = the real trade
  [ok] offset 60: 14:30 (102) -> 09:20 next day (104)

2. Real data: engine against the independent reference
  7067434 minutes 2004-06-11 00:19 .. 2026-10-07 16:57; candidate night trades 6827, day trades 5661
  [ok] GN-NIGHT-L: 5704 trades, reference 5704, identical
  [ok] GN-DAY-S: 5661 trades, reference 5661, identical
  [ok] GN-LS: 11365 trades, reference 11365, identical

3. Control at offset 0
  [ok] GN-NIGHT-L: 5704 control trades = 5704 real trades
  [ok] GN-DAY-S: 5661 control trades = 5661 real trades
  [ok] GN-LS: 11365 control trades = 11365 real trades

PASSED