PASS synthetic 1: events golden@3, death@6, golden@10
PASS synthetic 1 long-only: 2 trades, entry 14 -> 8 net -330.76, then open at end 12.25 -> 11.75 net -30.76
PASS synthetic 1 long-short: 3 trades (-330.76, short 8 -> 12 net -230.76, long open -30.76)
PASS synthetic 1: gross = price move only
PASS synthetic 1 gold: net points -6.30, percent -45.0, financing 3 days at 3%/yr
PASS synthetic 2 (first-state rule): events death@6, golden@10 only
PASS synthetic 2 long-only: one trade, opened on the first observed golden cross (14, open at end, net -30.76)
PASS synthetic 2 long-short: first position is the short opened by the death cross (9 -> 14, net -280.76), then long open -30.76
PASS roll cost: trade held across 2021-03-11 pays one extra $30.76 (net -361.52)
PASS roll dates: 2021 = Mar 11, Jun 10, Sep 9, Dec 9
PASS roll counting: entered the roll day session -> charged; exit fill on the roll day -> not charged
PASS worst open loss: spike of 5 points inside the trade = $250 (the exit bar, 6 points away, is excluded)
PASS EMA(3) of 1..5 = [NaN, NaN, 2, 3, 4]
PASS SMA(3) of 1..5 = [NaN, NaN, 2, 3, 4]
PASS state series: shift 0 reproduces the real events
Golden cross engine self-test. Run 2026-10-08T05:16:55.440Z
Real data (no results printed):
PASS ES: day keys strictly increasing, no duplicate day
PASS ES: daily bars identical to the independent construction (1742 days; 0 short days dropped)
PASS ES: 245-262 session days (counted days plus data-hole days) in every full calendar year - years 2021-2025
PASS ES: the fill minute directly follows the signal day's last bar (by construction i1+1 < n)
PASS ES GC-L-SMA: engine = independent reference, trade by trade
PASS ES GC-LS-SMA: engine = independent reference, trade by trade
PASS ES GC-L-EMA: engine = independent reference, trade by trade
PASS ES GC-LS-EMA: engine = independent reference, trade by trade
PASS ES: rolling SMA(50) and SMA(200) match a naive recomputation to 1e-9 - max diff 0.0e+0
PASS ES: rolling EMA(50) and EMA(200) match a naive recomputation to 1e-9 - max diff 0.0e+0
  (ES: days where the two averages are exactly equal after the warm-up: 0/0/0/0 for the four variants; the spec's >= / <= rule handles them)
PASS XAUUSD: day keys strictly increasing, no duplicate day
PASS XAUUSD: daily bars identical to the independent construction (5702 days; 34 short days dropped)
PASS XAUUSD: 245-262 session days (counted days plus data-hole days) in every full calendar year - years 2005-2025
  (XAUUSD: data holes, days with fewer than 30 one-minute bars, not counted as days: 2006-01-02; 2006-07-04; 2010-12-20 .. 2011-01-13; 2012-03-06 .. 2012-03-22)
PASS XAUUSD: the fill minute directly follows the signal day's last bar (by construction i1+1 < n)
PASS XAUUSD GC-L-SMA: engine = independent reference, trade by trade
PASS XAUUSD GC-LS-SMA: engine = independent reference, trade by trade
PASS XAUUSD GC-L-EMA: engine = independent reference, trade by trade
PASS XAUUSD GC-LS-EMA: engine = independent reference, trade by trade
PASS XAUUSD: rolling SMA(50) and SMA(200) match a naive recomputation to 1e-9 - max diff 1.0e-11
PASS XAUUSD: rolling EMA(50) and EMA(200) match a naive recomputation to 1e-9 - max diff 0.0e+0
  (XAUUSD: days where the two averages are exactly equal after the warm-up: 0/0/0/0 for the four variants; the spec's >= / <= rule handles them)

SELF-TEST PASSED: every check passed (4 s). ma_run.js may now be pointed at the data.
PASSED 2026-10-08T05:16:59.717Z