This robot wins 88.8% of its trades. That number is real, we counted it deal by deal. Here is the number that belongs next to it. The best single day in twenty years earned 12.95 USD and the worst single day lost 14.79 USD, so one bad day undoes more than one good day builds. A high win rate says how often a system is right. It says nothing about what happens when it is wrong, and in TwisterPro Scalper the difference between those two questions is the whole product.
One deterministic tester run of version 3.20 on XAUUSD# M15, 100% real ticks, first trade 2006-01-06 and last trade 2026-08-12. The offered window starts on 2003.05.05, so the first thirty two months of it carry no trade at all. 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 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 78.88 USD across 1,972 deals. With swap that is 19% of gross profit and our cost lamp is green. The final net is 329.72 USD from 986 trades.
The vendor's tester header reads well. Net profit 408.60 USD before our retrofit, profit factor 1.57, recovery factor 3.78, maximum balance drawdown 105.70 USD or 0.11%, and a Sharpe ratio of 14.53. That Sharpe is computed per trade on 986 tiny results against an account of 100,000 USD. We compute the same statistic on the end of day balance curve, the way a fund would, and get 0.74. Compound growth on the measurement frame is 0.02% per year. Both figures describe the same run. Only the second one compares to anything outside a tester.
The exit census explains it in one line. All 986 trades ended at a stop. Not one closed at a take profit, by signal or by time. Of those stop exits 876 closed in profit, which is the 88.8% win rate exactly. This is a trailing stop making every exit decision, and our detector confirms the trailing signature. A trailing stop that follows price closely converts most trades into small wins by construction, so the win rate is a property of the exit mechanism rather than evidence of forecasting skill.
The holding times show the same shape from the other side. The median trade is held 0.02 hours, about seventy seconds. The average is 1,996 seconds, roughly thirty three minutes, and the longest trade ran 98,021 seconds or more than a day. An average that sits twenty eight times above the median means a small number of trades stayed open far longer than the rest, and those are the ones that carry the losses.
Sort the record by year and one entry does not fit the story the win rate tells. In 2013 the system took 54 trades, won 74.1% of them, and finished the year at -72.09 USD. Forty winning trades out of fifty four, and the year still cost more than a fifth of everything the system earned across twenty years. The short side alone lost 50.18 USD in that year.
2013 is the extreme case and it is not the only one. Five of the twenty one years in the record are negative, which our audit rates caution. 2014 loses 13.02 USD at an 84.0% win rate and 2006 loses 8.23 USD at 66.7%. Every one of these years would have looked healthy on a win rate dashboard.
Now look at when the profit was made. 2025 delivered 73.69 USD from 116 trades and 2026 delivered 122.09 USD from 87 trades by 12 August. Together those two carry 195.78 USD of the 329.72 USD total, which is 59.4%. The nineteen years before them produced the remaining 133.94 USD. The single strongest year in the entire record is the current partial one, and it packs its 87 trades into the 224 days from 1 January to 12 August, a higher trade rate than any full year in the record reaches.
Concentration confirms the picture from the day side. 80% of the profit was made on 41 days, which is 4.4% of the 936 days with a closed trade. Remove the twenty best trades out of 986 and the net falls from 329.72 USD to 174.04 USD, so two percent of the trades carry 47% of the result. Between the first and the last trade our end of day series counts 5,373 weekdays, and the robot has a closed trade on 936 of them.
Our end of day analysis splits the curve into 24 drawdown episodes. The deepest peaked on 2012-07-16, bottomed on 2015-05-22 and did not recover until 2024-10-08. That is 3,191 trading days underwater for a depth of 114.74 USD, or 0.11% of the measurement frame. Percentage risk looks like nothing here because the position size is fixed at 0.01 lots against 100,000 USD. The shape is what survives the change of frame, and the shape says the equity curve went nowhere between 2012 and 2024.
The vendor recommends a deposit of 250 USD, so our capital replay uses that figure rather than a fallback. On that basis none of the 1,972 possible start points ends in a dead account, and the single observed path survives all 7,522 days of it. Re-expressed against 250 USD of capital, though, the same observed path carries a maximum drawdown of 45.9%, and all three prop firm floors we model are breached on the observed path itself rather than only in simulation. At 1,000 USD the drawdown falls to 11.47% and the floors are still breached. The first capital level at which no modelled floor breaks is 2,500 USD, ten times the recommended deposit, where drawdown reads 4.59%.
Our prop fit lamp is red and the reason is the consistency rule. Best day share exceeds the allowed percentage in 9 years under the funded profile we model. A system that makes 80% of its profit on 41 days conflicts with consistency rules by construction, whatever its win rate is.
Over the traded span the entry price series of this run rises from 619.24 to 4,250.79, a move of 586.45%. Holding 0.01 lots through those twenty years would have earned 3,631.55 USD. The EA earned 329.72 USD on the same lot size, which is under a tenth. That comparison is approximated from entry prices and holding gold means sitting through every gold crash since 2006, so it is not a fair risk comparison. It is a scale comparison, and scale is the first question a buyer should ask.
TwisterPro Scalper was published on 2026-02-24 and updated 20 times between 2026-02-26 and 2026-08-07, which is roughly one update every eight days. Our test window ends on 2026-08-12, so the stretch after the last update is 5 days and 2 trades worth 2.26 USD. That is not an out of sample test and we report it as one line rather than a result.
The two in sample windows are worth comparing. The 919 trades before release carry a profit factor of 1.35 and 239.82 USD across more than nineteen years. The 65 trades in the 165 days between release and the last update carry a profit factor of 4.23 and 87.64 USD. Both windows were inside the development period, and the later one is the shorter, denser and far more profitable of the two. That pattern is what one would expect from software still being changed, and it is exactly why we mark both windows as in sample.
One date counts in the product's favour. We measured version 3.20 and the market listing still shows version 3.20, so this audit describes the software that is currently on sale. That is not always the case, which is why every audit we publish names its version and its date.
The mechanics are conservative and we say so plainly. No martingale, no averaging into losses, never more than one position open, one symbol, 0 hedged moments, 0 failed entries and 0 failed modifications in the journal. Long and short are close to balanced at 483 and 503 trades, and the profit splits 181.72 USD long against 148.00 USD short. Our structure lamp is green. 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.