Logan made money in twelve of the thirteen years in which it was active. Over 23 years of gold ticks it traded 1,372 times on 381 trading days. 2025 made +$1,701.10. 2026 made +$2,295.79 with a win rate of 89.3%. The twelve profitable years add up to +$5,913.29.
The run finished at -$24,334.65 on a $100,000 account.
The gap between those two sentences is the audit.
The thirteenth year is 2016. It contains 20 trades, a win rate of 0.0%, and a result of -$30,247.94. The worst day table names it precisely: 2016-12-09, at -$30,247.94. The whole year is that day.
Put the two numbers side by side. Twelve years of work, +$5,913.29. One session, -$30,247.94. The single worst day is five times everything the system earned in more than two decades, and it is also the entire reported drawdown. The report head prints a maximum balance drawdown of 30.01% and a maximum equity drawdown of 32.09%. Our reconstruction from the deal timestamps agrees, at 30.06% closed and 32.11% floating.
The structure section is unambiguous. 285 entries were added against an open losing position and 0 with the move. Not most, not nearly all. All of them. Up to 20 positions were open at once on a single symbol. Long and short were held simultaneously in 23 moments. Volume rose after 28.8% of losses versus 4.5% after wins.
This is the averaging pattern, and its statistics are the ones you would predict. A strategy that adds to losers wins almost every trade, because every position that is allowed to run far enough gets closed at a small profit eventually. What it does not do is control the one case where the market keeps going. The exit mix confirms it. All 1,372 trades closed on signal or time and none on a stop loss. There was no stop in this run. When gold moved in December 2016 and did not come back within the basket's tolerance, nothing in the system was designed to say no.
Worth noting for fairness: the median size of an adverse add is 1.0 times the basket average, so this is averaging rather than classic doubling. That distinction matters for how fast the account dies. It does not change whether it dies.
The cost block carries a detail that is easy to skip. Commission across the whole run was -$167.54, retrofitted at $3.50 per lot per side, because the average position is only 0.016 lots and the trade count is low for a 23 year window. Swap was -$6,448.44, roughly 38 times the commission, across 216 overnight trades.
That is the price of holding baskets open and waiting for them to come back. It is invisible in any headline metric and it is more than the entire amount the strategy earned in its twelve good years. Gross before costs was -$17,718.67, so financing turned a bad result into a worse one rather than causing it, but on a version that holds longer or trades larger, this line scales with the waiting.
Our prop fit sweep runs the same trade path at reduced size against each rule set. Against IQ Capital Classic the account dies at full size, at 1/1.5, at 1/2, at 1/3 and at 1/4. The first size that survives is 1/6, and at that size the modelled withdrawal rate is 0.01% per year. At full size it would be 0.42% per year, which is already not a business.
The per position loss limit tells the same story from another angle. It is breached 15 times at full size and still 9 times at one third size. A funded account is hard breached on the second violation. And the consistency rule, which caps the best day's share of annual profit, is breached in 6 years. In 2008, to take the first, a single day carried 69.3% of the year.
The challenge model names the rule that does the damage. Against the FTMO rule set, phase one is passed on 0.0 of the 1,000 resampled paths and both phases together on 0.0. The paths do not run out of time either. 0.666 of them end on a daily loss day, and only the remaining 0.334 are still undecided when the data ends. The maximum loss floor is never what stops them, at 0.0. So the binding constraint here is the size of a single bad day rather than the depth of a slow decline, which is the same thing the per position loss limit says one paragraph above.
The measurement itself is as good as our protocol gets. 100% real ticks from 2003.05.05, 1,409 absent minutes out of 7,898,922 minute bars in the feed, 0 randomization prints, so the run is deterministic and reproducible. Log coverage is 1.0, chronology violations 0, duplicate deal ids 0, and the reconstructed final balance matches the report to the cent at $75,665.35. Whatever else is uncertain here, the arithmetic is not.
What the run cannot tell us is how a 23 year window with only 381 trading days behaves in general. The strategy sat out entire years. It is also worth saying plainly that gold rose 390.31% across the tested span, from 898.78 to 4,406.80, and the system still lost money while trading long 51.7% of the time.
This audit measures v2.95. Between 2026-08-20 and 2026-08-23 the vendor published versions 3.0 through 3.14, and the changelog entries name exactly the mechanics above. Grid steps capped at 5. A cascading recovery hedge from step 4. The internal drawdown limit lowered, twice. Two counter trend sub strategies removed.
We take that seriously enough to have queued a re-test, and we will not pretend to know its outcome. A cap on grid steps is the right category of fix for the failure this audit found. Whether the cap is low enough, and what the recovery hedge does in a December 2016, are empirical questions that need the same 23 years of ticks run again. Until that run exists, the honest statement is the narrow one. In the version we measured, the loss came from one day, the one day came from adding to losers without a stop, and the vendor has since changed that part of the machine.