Gold House places pending orders. Five sub strategies can be switched on, our deal census finds four of them in the trades, and together they opened 10,402 trades across 23 years of gold. The run ends at +$11,054.32 on a 100,000 USD measurement frame with a deepest reconstructed drawdown of 0.61%. Lots are fixed at 0.01 and stay fixed. Volume increase after a loss 0.0%, after a win 0.0%. There is no martingale here.
What makes this audit worth reading is not the total. It is the first four years.
The yearly table splits cleanly in two. 2003 loses $80.38 at a win rate of 18.9%. 2004 loses $222.24 at 21.2%. 2005 loses $168.79 at 25.2%. 2006 loses $65.17 at 33.7%. Those are the 4 negative years our regime dimension flags. Then 2007 turns positive at $67.79 and every single year from there to the end of the run is positive, with win rates from 2008 on spanning 43.6% to 65.2% at a median of 52.5%. The two largest years are the two most recent ones, 2025 at $2,085.95 and the partial 2026 at $1,353.16.
We are not going to call that proof of curve fitting, because it is not. Gold traded near 342 at the start of this run and near 4,408 at the end, and a pending order system built around modern gold ranges can fail on a market that moved a few dollars a day without anybody having tuned anything. But the shape is the shape. The years the developer could not have been optimising toward, because nobody optimises toward a market that no longer exists, are exactly the years that lose.
Gold House reached the MQL5 market on 2026-02-12 and was last updated on 2026-06-16, after 5 updates. Our release x-ray splits the run at those dates. Before release there are 8,318 days and 9,999 trades carrying +$10,426.91. Between release and last update there are 125 days and 259 trades carrying +$465.98. After the last update there are 57 days and 144 trades carrying +$161.43 at a profit factor of 1.37 and a win rate of 42.4%.
That last window is the only part of this backtest that was not available to whoever built the EA. It is positive, which is better than negative, and it is 57 days long, which decides nothing. Anyone who tells you a two month out of sample window validates a strategy is selling something.
Every trade comment starts with Gold House_Pending and ends in a letter, so the deals sort themselves. The four groups our engine finds add up exactly to the total result, which is a good sign that nothing was missed.
Group B earns $3.51 per trade while group E earns $0.85. One in twenty trades of this EA comes from the strategy that actually pays well, and it is the slow one. That is worth knowing before switching sub strategies off to reduce activity.
The positions do stack. Our structure section counts up to 4 open positions at once, 5,573 stacking entries which is 53.6% of all entries, and 2 moments with a long and a short open at the same time. Entries added in the direction of the move 813, entries added against an open losing position 136, which is 1.3%. Our engine rates this info rather than caution, and that is the right call. Four parallel pending strategies produce concurrent positions by design. A grid produces them by escalation, and the size escalation here is a flat 1.0.
Divide the result by the trades and you get $1.06, and our sensitivity section reaches the same number from the other side. The break even shock, the extra cost per trade that would flatten the whole run, is $1.06. The ladder shows what that means. An extra $0.50 per trade leaves +$5,853.32. An extra $1.00 leaves +$652.32. An extra $2.00 turns the audit into -$9,749.68.
The measured average position is 0.01 lots, so those shocks are not exotic. They are the difference between a good broker and a mediocre one. In our own cost model, commission across the run was -$832.16 retrofitted at $3.50 per lot per side, swap was -$140.79 across 411 overnight trades, and costs took 8.1% of gross profit, which our engine rates ok. A system that holds most trades for under half an hour barely pays financing. It pays spread, and spread is the one number a tester report cannot check for you.
Of the 10,402 trades, 10,178 ended on a stop loss, 44 on a take profit and 180 on a signal or a timeout. That alone would be unremarkable. The interesting part is that 4,923 of those stop exits closed in profit, which is 48.4% of them, and our engine therefore labels the run with a trailing stop signature.
The vendor default inputs say otherwise. Inp_TSLMode is 0, Inp_TSLStartPoints is 0 and Inp_TSLStepPoints is 0. Either the EA moves stops somewhere outside the trailing block that the input panel exposes, for example a break even move tied to the pending logic, or the label is picking up a different mechanism that produces the same fingerprint. A tester report cannot decide between those two, and we are not going to pretend it can. What a buyer can take from it is narrower and still useful. Almost every trade of this EA is closed by a stop order, so the quality of stop execution at your broker is not a detail of this strategy. It is the strategy.
80% of the profit was made in 212 days out of 3,192 trading days with trades, which is 6.6% of them. That earns a caution in our concentration dimension. The best single day made $164.30 and the worst lost $73.74, so this is not a jackpot profile. It is the ordinary shape of a system with a thin per trade edge, where a few good clusters carry the year.
Against the IQ Capital Classic rule set our sweep finds 0 account deaths at every size from full down to one eighth, and 0 breaches of the per position loss limit at every size. The floor is never touched. The verdict is still caution, and only because the consistency rule, which caps the best day at 30% of annual profit, is breached in exactly 1 year. That year is 2007, where the best day made $31.24 and the whole year made $67.79, so the best day is 46.1% of it. The rule is tripping over a year that earned almost nothing, not over a lucky Tuesday.
On the vendor's recommended deposit of 2,000 USD our fresh start map finds 0.0% of the 20,804 possible start points dying. The withdrawal rate at full size against this profile is 0.44% per year. Those figures come from 0.01 fixed lots on a 100,000 USD frame, which is a measurement setting and not a recommendation.
The challenge model looks friendly, and the time column takes most of it back. Over 1,000 resampled paths phase one is passed on 0.877, phase two on 1.0, and both together on 0.877, with a floor failure probability of 0.0. On that rate the audit expects 1.1 attempts and 2 attempts for a 90% chance of clearing phase one. Then comes the length. A path runs 6,071 days and the median passing path needs 5,457 of them to reach the phase one target of +10%. This profile passes because it almost never breaks a rule, not because it gets anywhere quickly, and at every sizing below full size the pass probability falls to 0.0, because the target stays fixed while the trading that has to reach it shrinks.
Our engine tracks what the traded instrument did over the same window. Gold went from 342.65 to 4,408.72 in the entry price series of this run, a move of 1,186.64%. Buying 0.01 lots at the start and holding to the end would have made $4,066.07. The EA made $11,054.32, so it clears that bar by a factor of more than two, and it does so with a reconstructed drawdown of 0.61% instead of every gold crash of the last two decades. That is a real answer and it deserves to be said plainly, because most of the EAs we take apart cannot claim it.
The run is technically clean. 100% real ticks from 2003.05.05, 1,409 absent minutes out of 7,899,407 minute bars, 0 failed entries, 0 failed modify attempts and 0 randomizer prints, so the run is deterministic and can be repeated. Log coverage is 1.0, chronology violations 0, duplicate deal ids 0, and the reconstructed final balance matches to the cent at $111,054.32. Trade pairing is exact and taken from the journal rather than inferred.