Of the 28 EAs in our catalog, 22 trade gold. OilVector X trades WTI crude oil and it is the only one that does. That alone made it worth measuring, because every structural habit we have learned to look for was learned on a different instrument.
The run opened 1,192 trades over 8.9 effective years and closed at +$35.23 after commission, on a 100,000 USD measurement frame with a fixed 0.01 lots. Three of our six verdict dimensions come out red. Costs, concentration and prop fit.
The architecture is not the reason. The size of the result is.
Our engine tracks what the traded instrument itself did over the same window. In the entry price series of this run WTI went from 50.826 to 84.532 across 2017-10-05 to 2026-08-20, a move of 66.32%. Buying 0.01 lots at the start and holding them to the end would have produced $33.71 on the same measurement frame. The EA produced $35.23.
That is a difference of $1.52 for 1,192 trades spread over nearly nine years. And the comparison is not a rhetorical trick. All 1,192 trades were long, 0 were short, and 100% of the profit came from the long side. Our engine raises a long beta flag for exactly this shape, because a system that only ever buys a rising instrument is hard to distinguish from the instrument.
What the comparison leaves out belongs next to it. Holding crude oil from 2017 to 2026 means sitting through everything that happened to oil in 2020, while this run finished with a deepest end of day drawdown of 0.03% on the 100,000 USD frame. Per unit of pain the EA is far ahead. On the question a buyer actually asks, which is how much money the thing makes, nine years of trading added $1.52 over doing nothing.
Divide the result by the trade count and the audit fits into one number. $35.23 over 1,192 trades is about three cents each, and our sensitivity section arrives at the same figure from the other direction. The break even shock, the extra cost per trade that would take the net result to zero, is $0.03.
The cost shock ladder shows what that means. At an extra $0.50 per trade the run nets -$560.77. At $1.00 per trade it is -$1,156.77. At $2.00 it is -$2,348.77. The measured average position is 0.01 lots, so none of those shocks is exotic. They are the difference between our broker and a slightly worse one.
The same fragility shows up when we remove the best trades. Take out the 10 most profitable ones and $35.23 becomes $20.32. Take out 20 and it becomes $5.89. Twenty trades out of 1,192 stand between this result and nothing.
80% of the profit was made on 8 trading days. That is 0.8% of the 978 days on which this EA traded at all, and those 978 days are themselves only 42.06% of the 2,325 trading days in the window. The rest of the time it either sat still or moved money from one pocket to the other.
The daily extremes make the same point without any statistics. The best day was +$4.32 on 2019-04-23. The worst day was -$6.48 on 2023-11-03. The worst single day is larger than the best single day, and it is larger than a sixth of everything the run earned in nine years.
The yearly table is where this becomes a buying decision. 2022 made $22.83 of the total $35.23. Four years are negative, and the two most recent full ones are the worst on record at -$13.60 in 2024 and -$13.20 in 2025, before 2026 turned back up with +$18.63. Remove one good year and the picture is not thinner, it is gone.
Against that, the instrument itself moved +53.73% in 2021 while the EA earned $9.44, and +2.24% in 2022 while the EA had its best year. So this is not simply a long oil position in disguise either. The good year and the good market are not the same year.
There is no martingale here and our numbers say so plainly. Volume increase after a loss 0.0%, after a win 0.0%, median size escalation 1.0. The lot size is fixed and stays fixed. Moments with a long and a short open at once 0. One symbol, never more.
It does stack. Up to 4 positions were open at the same time, all of them long, across 315 stacking entries, which is 26.4% of all entries. Of those, 287 were added with the move and only 27 against an open losing position, a share of 2.3%. That is a scaling in behaviour, not a recovery grid, and our structure lamp reads info rather than red.
The exit mix is where the personality sits. 846 trades ended on a stop loss, 164 on a take profit and 182 on a signal or time exit. Only 2 of those stop exits closed in profit, so this is not a trailing stop signature. The win rate is 19.7% and the median holding time is 3.61 hours. Roughly one trade in five works, the winners are bigger than the losers, and the whole construction ends up three cents ahead per attempt.
This looks wrong at first glance. Our prop fit sweep against IQ Capital Classic 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 red, because the consistency rule, which caps the best day's share of annual profit at 30%, is breached in 4 of 10 years.
Our replay takes out everything above the start balance on the last trading day of each month. Over 107 months it paid in 15. Total withdrawn $50.13, median paid month $3.06, best month $6.76, longest dry streak 38 months. Three years without a payout, inside a run that our verdict does not call a failure.
And the number that follows from it. After those withdrawals the account ends at $99,985.20, which is below where it started. Sweeping the peaks and then living through the drawdowns leaves less than the sum suggests, and that is a property of the profile rather than an accident of the rule.
Nothing here kills an account. Our fresh start map on the vendor recommended 1,000 USD deposit finds 0.0% of 2,384 possible start points dying. Nothing here fills one either.
The run is clean by our protocol. 100% real ticks from 2017.10.02, 209,721 bars reported at a bar coverage of 98.05%, 0 failed entries, 0 failed modifies and 0 randomizer prints, so the run is deterministic and reproducible. Log coverage is 1.0, chronology violations 0, duplicate deal ids 0, and trade pairing confidence is exact because it comes from the journal rather than from inference. The first entry landed 2017-10-05 17:20:21, three days after the window opened.
One number pair is worth carrying away. The end of day Sharpe of this run is 0.38. The tester head prints 2.83 for the same run. Same trades, two different ways of asking how steady they were, and the gap between them is the reason we do not quote report heads.
What the measurement cannot tell us is whether three cents per trade survives contact with a live crude oil spread. That is exactly the quantity a tester report is worst at.