This robot wins 49.9% of its trades. A coin flip, and 420 of its 622 stop exits close in loss. Here that is not the end of the story. Across 945 trades between 2003-10-03 and 2026-08-10 the account still finishes 1,271.79 USD ahead on our measurement frame, because the average trade returns 1.43 USD and the winners are bigger than the losers. That is what a breakout system is supposed to look like. It also has three problems that a win rate never shows, and one of them is that the version we measured is no longer the version on sale.
One deterministic tester run of version 7.0 on XAUUSD# on the Daily chart, 100% real ticks, first trade 2003-10-03 and last trade 2026-08-10. We reconstruct the run deal by deal against the tester journal, so pairing confidence is exact, journal coverage is 1.0, chronology violations 0, duplicate deal ids 0, and the reconstructed balance matches the reported one to the cent. The journal records 0 failed entries and 0 failed modifications. The catalog protocol uses a fixed 0.01 lot on a 100,000 USD frame at 1:5000, which is a measurement setting and not a recommendation. On top of the tester result we retrofit commission at 3.50 USD per lot per side, which costs this run 75.60 USD across 1,890 deals.
The vendor's tester header reads net profit 1,347.39 USD before our retrofit, profit factor 1.54, recovery factor 17.31, maximum balance drawdown 70.00 USD or 0.07%, and a Sharpe ratio of 2.85. We recompute Sharpe on the end of day balance curve, the way a fund would, and get 1.01. Compound growth on the measurement frame is 0.05% per year. The gap between those two Sharpe figures is 1.84 points here, and in the widest case we have measured in this catalog it is 72.06 points. The reason is mechanical rather than flattering. A system that holds a position for hours or days has per trade and per day statistics that live at similar resolutions, so there is less room for the aggregation gap to open.
The exit census is the most informative table in this audit. 622 trades ended at a stop, 271 at a take profit and 52 by signal or time. Of the stop exits only 202 closed in profit, which is 32.5%. Our detector finds no trailing stop signature, so a stop here usually means the trade was wrong and got cut, not that a profit was locked in.
That is the honest shape of a breakout strategy. The failures are cut at a fixed distance, and the few that run pay for all of them. The holding times confirm it. The median trade is held 5.33 hours, the average 68,348 seconds which is about nineteen hours, and the longest single trade ran 1,642,994 seconds or more than nineteen days. Expected payoff per trade is 1.43 USD against a 49.9% win rate, and both numbers are needed before either one means anything.
Hold trades that long and the financing bill changes character. Our cost decomposition finds commission of 75.60 USD and swap of 173.39 USD, so the swap line is more than twice the commission line. The audit raises a specific note for this, because it is unusual. 363 trades were held overnight out of 945. Together the two lines take 16.4% of gross profit, which is a healthy ratio, and our cost lamp carries an information flag rather than a warning.
This matters for anyone comparing brokers, and the two numbers say how much. Financing costs this run 0.18 USD per trade and commission costs it 0.08 USD per trade, against an edge that survives up to 1.35 USD per trade of extra cost. So a commission rate that is half a dollar per lot worse than ours moves this result by about a cent per trade, while the financing rate moves it more than twice as hard as the commission rate does. Neither rate is something our tester run can know about the account you would use, which is why the figure to look up before running this is the gold swap of that account and not its commission table.
We test that directly instead of guessing. Charge an additional flat 0.50 USD per trade on top of everything already counted and the run still finishes at 799.29 USD with a profit factor of 1.286. Charge 1.00 USD per trade and it finishes at 326.79 USD with a profit factor of 1.107. The edge disappears at 1.35 USD per trade, and at 2.00 USD the run turns into a loss of 618.21 USD.
Read that against the average lot size in this run, which is 0.01. The reserve of 1.35 USD is the net result divided by the trade count, 1,271.79 USD over 945 trades, so it is the same edge seen from the cost side rather than a second finding. What makes it worth stating is the size of the position it belongs to. A one cent lot on gold carries that much room per trade, which is why the swap question above is worth checking rather than assuming.
Every one of the 24 calendar years in the record carries trades, between roughly thirty and fifty of them per year apart from the partial first year, and five of those years are negative. Four of the five sit early, in 2003, 2004, 2005 and 2007, with 2005 the worst at -35.75 USD. The only later negative year is 2017 at -5.97 USD. Our audit rates that caution, and the count is not mild by our own standard either. Counted on 2026-08-27, 16 of the other 24 audits in our catalog carry fewer negative years than this one and only six carry more, so the reassuring part of this record is where the losing years sit rather than how many there are.
Concentration tells the same story. 80% of the profit was made on 74 days, which is 7.9% of the 936 days with a closed trade. Remove the twenty best trades out of 945 and the net falls from 1,271.79 USD to 965.15 USD, so two percent of the trades carry 24% of the result. The current partial year 2026 carries 174.20 USD, which is 13.7% of the total. Those three figures describe a system whose result does not hang on a handful of months. It still hangs on a minority of days, which is why the concentration lamp is caution and not green.
Our end of day analysis splits the curve into 95 drawdown episodes. The deepest peaked on 2007-01-23, bottomed on 2007-08-10 and recovered on 2008-01-09. That is 251 trading days underwater for a depth of 71.92 USD, or 0.07% of the measurement frame, across 24 trades of which 19 lost. The deepest hole in twenty three years lasted under a year and was made of two dozen trades. Percentage risk looks tiny here only because the position size is fixed at 0.01 lots against 100,000 USD, so read the shape rather than the percentage.
The vendor names no deposit recommendation, so our capital replay falls back to 1,000 USD. That figure is ours, not the vendor's, and everything below depends on it. Re-expressed against 1,000 USD the observed path carries a maximum drawdown of 7.19%. Two of the three prop firm floors we model are breached on the observed path, the IQ Capital profile and a generic 6% trailing rule, while the FTMO challenge floor is not breached on the observed path although our simulation puts its probability at 0.54. Move to 2,500 USD and drawdown falls to 2.88% with no floor broken on the observed path at all. None of the 1,890 possible start points ends in a dead account, and the observed path survives all 8,347 days of it.
Our prop fit lamp is still red, and the reason is the consistency rule. Best day share exceeds the allowed percentage in 8 years under the funded profile we model. A system that makes 80% of its profit on 74 days out of 936 conflicts with consistency rules by construction, even when its drawdown behaviour is mild.
Over the traded span the entry price series of this run rises from 378.29 to 4,101.18, a move of 984.14%. Holding 0.01 lots through those twenty three years would have earned 3,722.89 USD. The EA earned 1,271.79 USD on the same lot size, which is about a third. The comparison is approximated from entry prices and buying and holding gold means sitting through every gold crash since 2003, so it is not a risk adjusted comparison. It is a scale comparison, and what it says is that twenty three years of signals did not beat the metal itself. Note also that the short side earned more than the long side here, 704.53 USD from 428 short trades against 567.26 USD from 517 long trades, so the result is not a disguised long gold position.
Now the part that limits everything above. We measured version 7.0. Ultimate Breakout System was published on 2025-05-16 and has been updated 49 times between 2025-05-23 and 2026-08-23, which is roughly one update every nine days. Our market watch on 2026-08-27 recorded the listing at version 7.4.
The last update landed on 2026-08-23 and our test window ends before that, so the out of sample window of this run is 0 days and 0 trades. Not small. Zero. Every trade in this audit was produced by software that was still being changed. The two in sample windows also differ sharply. The 893 trades before release carry a profit factor of 1.42 and 1,043.02 USD across more than twenty one years. The 52 trades in the 452 days between release and the last update carry a profit factor of 4.35 and 228.77 USD. A development period that looks three times better than the two decades before it is the normal signature of software being tuned, and it is why we mark both windows as in sample.
One more thing belongs here. The input panel of version 7.0 has 236 entries, of which 19 are section headers with an empty value. More than two hundred settings is a very large surface to tune against a history, and it is the reason a version specific, dated audit matters more for this product than for a robot with twenty knobs.
No martingale, no averaging into losses, 0 adverse additions to a losing position, one symbol. Two structural notes rather than warnings. The system holds up to 2 positions at once, and there are 10 moments in twenty three years with a long and a short open at the same time. Both read as parallel sub strategies rather than as a grid, and our structure lamp carries information flags, not a caution. The red lamp on this product is prop fit and the caution lamps are concentration and regime.
The deal level census, the drawdown decomposition and the capital replay need full deal reconstruction, which 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.