Forward Test: Gold Golden Cross, Second Test (GH-L-N-2)

01Why there is a second test

• The actual (demo) series of GH-L-N was stopped on 15 October 2026: trades unrelated to the rule, placed on the same account, triggered its kill limit, and its only trade already carried deviations. Its protocol does not allow the limit to be reset or the test to be resumed: a restart is a separate test (GH-L-N, sections 5 and 10). This is that separate test.
• What changes (everything else is GH-L-N unchanged): the series is made only of this test's own orders, and its kill limit is measured on its own result, not on the account (sections 4 and 6); the owner may trade other things on the same account, which is reported; and the first signal is set in section 3.
• GH-L-N's paper series continues and is the simulated reference for both tests. The two tests are never combined into one record.

02The rule (frozen; the same as GH-L-N)

• Market: XAUUSD at IC Markets, the same broker and symbol as the historical data. H1 bars, New York clock (the broker's server time is New York + 7 hours), bar time is the close time. SMA of the closes, fast 50 and slow 200.
• Golden cross at the close of bar k: SMA50[k] > SMA200[k] and SMA50[k-1] <= SMA200[k-1]. Death cross: the reverse.
• Entry: buy at market as soon as possible after a golden-cross bar closes. Exit: close at market as soon as possible after a death-cross bar closes.
• Long only. One position at a time. No stop loss, no take profit, no scaling, no re-entry in the same state, no filters, no discretion. All sessions. Positions are held overnight and over weekends and holidays.
• R as in the historical test: 1.5 x ATR(14) of the H1 bars at the signal bar, in dollars per ounce; a normalisation unit, not a maximum loss.

03The first signal

• Eligible signals: crosses whose signal bar closes at or after 16 October 2026, 17:00 New York, the close of the last bar of the week before the start.
• This protocol was written after that bar closed and before any trade on it could be entered (the market was closed for the weekend). On a second price feed, the averages crossed upward on that bar by $0.18, a golden cross; the broker's MetaTrader 4 terminal had not yet written that hour's bars when this was written. The program reads the cross from the broker's MetaTrader 4 1-hour bars. If they show the golden cross on that bar, the first position is bought at the start; if they do not, no position is opened until the next golden cross. The outcome of that trade was not known when this was written.

04Account, size and the two series

• Account: the same IC Markets Standard MT4 demo account in USD as GH-L-N (the owner cannot open a second demo account). The owner may trade other things on this account; those trades are not part of this test, and the account's balance and equity are reported only for information. Account identifiers are kept private. A balance of at least $10,000 is recommended: with no stop, one ounce could be several hundred dollars against the position at its worst (GH-L-N, section 3).
• Size: a constant one ounce of XAUUSD (0.01 lot with a 100-ounce contract).
• The actual series is made only of the orders that carry this test's identifier (magic number 20261018), placed by the program GH_L_N2_GoldForward. Its result comes from the actual demo fills and charges: the spread is inside the fills and never subtracted again; commission and swap are added as charged.
• If other trading on the account closes or blocks this test's position (a margin call, a stop-out, or not enough free margin for the order), it is recorded as an incident and a deviation. A missing order for lack of margin is recorded as a missed signal.
• Paper series: the GH-L-N paper series (fills at the next 1-minute open of the broker's history, a flat $0.30 an ounce, no financing) is the simulated reference, labelled simulated everywhere.

05Running the program

• The program checks the rule on every price tick and every 10 seconds, and must run without interruption (MetaTrader 4 open, connected, AutoTrading on). Times when it was not running are recorded from its log.
• Late or missed fills: lateness is measured from the reference fill time, the next 1-minute open of the history after the signal bar (so a signal on the last bar of the week is not late at the next weekly open). A fill within 4 hours of the reference is taken at the actual price and flagged late; beyond 4 hours it is flagged missed, still counted as a signal, and results are shown with and without it.
• Only one copy of the program may run on the account.

06Kill limit (on this test's own result)

• The series' marked-to-market result is the net result of its closed trades plus the floating result of its open position, commission and swap included. If it falls $2,000 below its highest value since the start, the position is closed at market and further orders are suspended. The trigger time, the result, the execution and the costs are recorded.
• The limit is not reset and the test is not resumed. A restart would be a third, separate test. Other trading on the account does not count toward this limit.

07The journal (public, weekly)

• Published every Saturday (UTC) on this page, including weeks with no trades, with the fields of GH-L-N section 6: per signal, per trade (requested and actual fill, bid and ask, slippage against the requested price and against the reference, commission, swap per night, holding time, result, worst unrealised loss), daily marks at 17:00 New York of the open position, results in dollars per ounce, in R and in percent, and the closed-trade and marked-to-market drawdowns of the series.
• Deviations: any manual action on this test's position, any interruption of the program at a signal, and any effect of other trading on the position are logged with their reason. If more than 10% of the trades carry a deviation, the test is invalid and is reported as such.
• Costs are compared with the two historical thresholds at every update, as in GH-L-N section 7 ($0.66 an ounce per round trip of execution cost, 1.65% a year of financing).

08Evaluation criteria (fixed now; the same as GH-L-N)

• Formal evaluations at 50, 100 and 150 closed trades of this test's actual series. At about 17 trades a year, the first could take about three years. Weekly updates before it are descriptive only.
• Primary measure: the mean net R per trade with actual fills and charges, with its 95% confidence interval; Contradicted if the mean is zero or negative or the profit factor is below 1.0; Not contradicted if the mean is positive, the profit factor at least 1.0 and the interval still includes zero; Supported only if the lower end of the interval is above zero.
• Matched control: GH-L-N's circularly shifted state-series control on this test's window, 500 runs, seeds 1000 plus the run number, descriptive only.
• If the kill limit ends the series before 50 closed trades, the series is reported as stopped and not evaluated. The criteria are not changed after results are seen.

09Modifications

Any change to the rule, the averages, the timeframe, the symbol, the broker, the account, the size, the stop, the targets, the filters, the kill limit or these criteria is a separate test, with its own name, protocol and start date. Results are never merged across tests.

JWeekly journal

Updated every Saturday (UTC), including weeks with no trades. Actual demo results and simulated paper results are always shown separately; account identifiers are private. Dollars are per ounce unless stated.

No entries yet. The test starts on 18 October 2026 at 18:00 New York (22:00 UTC); the first weekly update is due on Saturday 24 October 2026.

Raw files

CCorrections

None yet. Any change to this protocol after the start would be published here and would make the changed version a separate test (section 9).