CME gap self-test, 2026-10-21T15:45:27.728Z

1. Hand-worked weekends
A. Friday 3 Jan 2020 close 7000; Sunday 5 Jan 18:00 New York (23:00 UTC) opens at 7140, a +2% gap -> short toward 7000; Monday 01:00 UTC low 6990
  [ok] gap +2.00%, filled 2 h 1 min after the reopen (to the close of the 01:00 minute)
  [ok] CG-S1: target, net 0.9495 R (R = the gap, 140)
  [ok] CG-S2: R = 2 x the gap; CG-NS: the same exit as CG-S1
  [ok] reference identical (3 variants)
B. same gap; Monday 00:10 UTC opens 7200 and trades up to 7300: CG-S1 stopped at 7280; CG-S2 (stop 7420) holds and reaches the target on Tuesday
  [ok] CG-S1 stopped at 7280: net -1.0515 R
  [ok] CG-S2 survives and fills on Tuesday
  [ok] reference identical
C. no fill all week: out at the open of the minute starting next Friday 17:00 New York (22:00 UTC) at 7150; a 0.5% gap the week after is measured but not traded
  [ok] CG-NS: time exit at 7150, net -0.1225 R
  [ok] second weekend: gap 0.50%, below 1%, not traded
  [ok] reference identical
D. clock change: Friday 6 Mar 2020 16:59 EST = 21:59 UTC; Sunday 8 Mar 18:00 EDT = 22:00 UTC (gap -1.5% -> long); next Friday 13 Mar 17:00 EDT = 21:00 UTC
  [ok] gap -1.50%, long, out at 21:00 UTC (8960), not 22:00
  [ok] reference identical

2. Real data: engine against the independent reference
  4789278 minutes 2017-08-17 04:00 .. 2026-09-30 23:59 UTC; 458 weekends measured, 250 with a gap of 1% or more
  [ok] gaps identical (458 / 458)
  [ok] CG-S1: 250 trades, reference 250, identical
  [ok] CG-S2: 250 trades, reference 250, identical
  [ok] CG-NS: 250 trades, reference 250, identical

3. Control at offset 0
  [ok] CG-S1: 250 control trades at the real entry minutes, same direction and R
  [ok] CG-S2: 250 control trades at the real entry minutes, same direction and R
  [ok] CG-NS: 250 control trades at the real entry minutes, same direction and R

PASSED