An expert advisor does not always get what it asks for. It sends an order, the platform refuses it, and the journal writes one line about it. Our engine counts those lines and every audit carries the result in two numbers, failed entry attempts and failed modify attempts. Of our 33 audit runs, 12 record at least one refused entry and 12 record at least one refused modification. Five audits carry a red flag because of it. This article is about what those lines really contain, because the number on its own says less than it looks like it does.
The flag fires in five audits, which are four distinct runs because our Quantum Queen and Quantum Queen X entries carry the same measured numbers throughout. Together those four runs carry 1,792 refused order lines. Quantum Titan holds 1,260 of them, Quantum Athena X 354, Gold Atlas 119 and Quantum Queen X 59.
We read the journals line by line. All 1,792 carry the same rejection reason, and it is not slippage, not a bad price and not a margin call. The Gold Atlas journal writes it like this at 2003.05.28 00:56:08, failed market sell 0.01 XAUUSD sl: 366.55 [Market closed].
The market was closed. The strategy asked outside trading hours and the platform declined to act on it. Not one of the 1,792 lines says the account ran out of money. That is worth stating plainly, because a count of refused orders reads like a solvency warning and in these four runs it is nothing of the kind.
Our counter matches on the journal phrase for a refused market order. A market order in that log can be an entry or a close, and the close carries an extra clause naming the ticket it wants to shut down. Counted separately, 853 of the 1,792 lines are close requests, not entries. In the Quantum Titan run alone, 728 of 1,260 read failed market sell 0.01 XAUUSD, close #584 buy 0.01 XAUUSD 1899.72 [Market closed], which is an attempt to shut down an existing ticket.
That changes the meaning rather than the size. A refused entry costs a trade that was never taken. A refused close leaves a position open that the strategy wanted flat, and it stays open until the session comes back. For a reader deciding whether to run something over a weekend, the second case is the more interesting one, and until now our audits folded it into the first.
The correction does not move any verdict. Our rule raises the flag when refused entries exceed 0.5% of the trade count. Counting only true entries, Quantum Titan sits at 532 against 1,947 trades, which is 27.3%, Quantum Athena X at 267 against 3,976, which is 6.7%, and Quantum Queen X at 21 against 3,357, which is 0.63%. Gold Atlas has no close requests at all, so its 119 stand unchanged at 1.1%. All four stay flagged. Our own Quantum Titan review repeats the uncorrected figure of 1,260, and this paragraph is the correction.
The second and third largest entry counters in the catalog belong to audits with no data quality flag at all. Scalping Robot Pro records 796 refused entries in the 4.0 run and 1,111 in the 3.0 run. Both runs also record more than 300,000 trades, so the shares are 0.26% and 0.37% and stay under the rule. A high scalping frequency buys a large denominator.
The lesson generalises past our own thresholds. A rejection count without the trade count next to it is unreadable, and so is a rejection count without its reason. Both belong on the page, and today only the count does.
The second number has no rule attached to it. Our red flag logic reads the entry counter and never looks at the modify counter, so an audit can publish a very large number there and every lamp stays green. Smart Gold Impulse records 2,810 refused modifications against 4,163 trades. CryonX records 1,414 against 4,127 and Lizard 908 against 21,524. Across all 33 runs the modify counters add up to 7,437 refused requests that no rule of ours reads.
Read individually they split into two very different habits. In Smart Gold Impulse, 2,805 of the 2,810 say [Market closed], and the timestamps show the shape. On 2011.10.04 the journal carries failed modify #953 buy 0.01 XAUUSD sl: 1610.11 to sl: 1623.12 [Market closed] at 23:59:23, then the same ticket again at 23:59:25, twice at 23:59:32, and again at 23:59:34 and 23:59:37, each request naming a slightly different stop between 1623.11 and 1623.14.
That is a trailing stop chasing a price into the daily break and retrying every few seconds on the same ticket. It costs nothing in the test. It is still a description of how the code behaves when a venue stops answering.
The other habit looks different. In Lizard, 906 of the 908 refusals say [Invalid stops], and in Adaptive Gold Scalper the same reason dominates its 192. The lines we sampled there ask for the values the position already carries. At 2023.12.22 01:05:30 the journal reads failed modify #8524 buy 0.01 XAUUSD sl: 2048.20, tp: 2148.04 to sl: 2048.20, tp: 2148.04 [Invalid stops], with the same two prices on both sides of the request.
The request is a no operation. The platform rejects it, the strategy repeats it, and neither side gains anything. Our CryonX review reads that EA's 1,414 refusals as a broker declining a stop placed too close to price. The journal disagrees with our review, 1,413 of the 1,414 say [Market closed], and the closed session is the mechanism there.
Three questions turn the number into information. Is it an entry or a close, because one costs a trade and the other leaves a position open. What reason does the journal give, because a closed market, an invalid stop and an empty account are three unrelated problems that share one counter. And how many trades sit underneath it, because 796 refusals across 302,798 trades and 1,260 across 1,947 are not the same finding in any sense.
None of this was hidden. The counters have been on every audit page since we started publishing them, the journals sit in our report database, and the arithmetic above is grep and division. It took reading the lines instead of the totals.
Every verdict lamp in the catalog is public on the audit pages. Your own tester report? The browser check is free.