This EA is profitable across the whole gold history we hold, and it still fails the rule set it is named after. Our engine dissected a strategy tester run of Prop Firm Gold EA v2.5 over the full available gold tick history. The run is profitable, +7,532.83 USD before our commission retrofit on a 100,000 USD measurement frame, with a reported maximum balance drawdown of 0.65% and a profit factor of 1.29 across 9,538 trades. No martingale, no hedging, no basket rescue. And then the prop rule replay puts the withdrawal rate at 0.28% a year and the consistency rule in breach in 7 years. Both statements come from the same run.
One tester run on XAUUSD, M30, from 2003-05-05 to 2026-08-16, history rated 100% real ticks, execution delay 10 ms parsed from the tester journal, fixed 0.01 lots on a 100,000 USD frame at 1:5000. Trade pairing came from the journal with exact confidence and 0 chronology violations, and our reconstruction reproduces the closing balance to 0.00 USD. Two structural findings are simply absent here. Lot size after a loss does not escalate (median ratio 1.0), and the run never holds long and short at the same time (0 hedged moments). Up to 3 positions run in parallel, which the audit rates as information rather than a warning.
The tester journal contains 6,030 randomization prints. The EA randomises its own trade decisions, so two runs of the same build on the same history do not produce the same trades. That is a vendor choice, not a defect, and the audit rates it as information. It does change what a published backtest can mean. A single run of a randomising EA is one draw from a distribution, and the marketed curve is the draw the vendor kept. Every number in this article therefore describes one specific run, identified by its report checksum in the audit, and not a repeatable property of the product.
The concentration flag is red. 80% of the profit was made in 67 days, which is 1.2% of the 5,533 trading days in the run. The best single day made 234.81 USD and the worst lost 108.57 USD, so the profile is a long grind punctuated by a handful of days that carry the result. Removing the 20 best trades takes 2,123.79 USD off the net, which is 6.4% of all gross wins from 0.2% of the trades. An edge that thin across time is not automatically fake. It is fragile, and prop rules punish fragility of exactly this shape.
We replay the audited history against published prop firm rule sets. Under the IQ Capital Classic profile (6% end of day trailing drawdown, 30% consistency, rules as of 2026-08-13) the account never dies, at any tested sizing from full size down to 1/8. That is the good half. The other half is the payout. At full size the sweep leaves a withdrawal rate of 0.28% a year, at 1/8 it is 0.03%, and the consistency rule is breached in 7 years because those few big days exceed the allowed share of annual profit. An EA that survives the drawdown floor and still cannot pass the consistency clause is a specific, checkable finding, and it deserves attention precisely because the product is aimed at prop traders.
The audit also models the FTMO Challenge, which asks a different question than the funded profile. Across 1,000 resampled paths this account never loses the challenge. The probability of hitting the maximum loss floor is 0.0 and the probability of hitting a daily loss limit is 0.0. It also almost never wins it. Phase one asks for +10% and is reached on 0.006 of paths, phase two asks for +5% and is reached on 0.941, and the combined figure is 0.006. The reason is time. Each simulated path runs 6,069 days, the length of the audited daily series, and the median path that does reach the phase one target needs 5,922 of them. On 0.994 of paths the history runs out first. At that pass rate the audit puts the expected number of phase one attempts at 166.7, and 383 attempts for a 90% chance of clearing phase one once.
Sizing down does not help, and the audit shows why. At every step below full size the phase one pass probability is 0.0, because the target stays fixed at 10% of the account while the trading that has to reach it shrinks. Read next to the funded profile this is consistent rather than contradictory. A low return engine with a small floor risk survives the rules that kill fast systems and misses the rules that ask for profit inside a deadline.
The product page records a release on 2025-10-29 and 15 updates, the last on 2026-03-28. Of the tested history, 8,209 days and 9,261 trades lie before the release (net +5,610.01 USD, profit factor 1.25). The window that is guaranteed out of sample, after the last vendor update, covers 138 days and 132 trades for +107.08 USD at a profit factor of 1.05. The middle window, from the release to the last update, ran at a profit factor of 1.57 over 145 trades, so this is not a straight decline across the three windows. What is true is that the only window the vendor could not see while building sits closest to 1.0. With 132 trades that window is too small to settle anything, which is the honest reading of it.
Gross profit is 7,532.83 USD. Our commission retrofit of 3.50 USD per lot per side over 19,076 deals costs 763.04 USD, or 10.1% of gross, leaving 6,769.79 USD. Swap is 0.00 USD because 0 trades stay open overnight, which is unusual and worth noting for anyone comparing this to a grid. The margin against a worse cost world is 0.71 USD per trade. At a flat shock of 1.00 USD per trade the run turns negative at -2,768.21 USD. A broker with wider gold spreads than our measurement world is not a detail for this profile, it is the whole question.
Across 1,000 bootstrap paths the probability of a net loss is 0.0, the median profit factor is 1.257 and the median expectancy is 0.71 USD per trade. The block bootstrap adds the number that matters for anyone waiting on a payout. The median time under water is 514 trading days, and the 95th percentile is 1,009. On the observed run the annualised return is 0.27% with an end of day Sharpe of 1.18. This is a low return, low drawdown profile, and it needs to be judged as one rather than against the returns a grid advertises.
The prop rule replays, the survival map and the deal level census need trade reconstruction, and that is what we sell. Any catalog audit is $19, and EAs with several audited presets or versions come as one bundle. Your own tester report? The browser check is free.