Our watch registered 98 expert advisors from the MQL5 new arrivals page between 2026-08-24 and 2026-08-27. 85 of them showed a price and the middle one asks 99.99 USD. Nine ask nothing, the dearest asks 990.00 USD, and 13 cards carried no price we could read. None of the 98 has been audited here.
Every night the watch spends one extra request on page one of the new arrivals listing. Each product card yields four things, the numeric product id, the name, the author and the advertised price. New ids are appended to a register that never forgets an id it has seen. Everything below is counted out of that register and out of the alert file the same run writes.
The first run is the exception and it matters for every count that follows. It had no register to compare against, so it stored the page as it stood and wrote a single operations line instead of 70 separate alerts. Those 70 entries carry the date 2026-08-24 because that is when we first looked, not because 70 products appeared that day.
After the seeding the register grew by 28 entries. One arrived on 2026-08-25, sixteen on 2026-08-26, eleven on 2026-08-27. The single entry on the first of those days has a mechanical reason rather than a market one. The seeding ran at 19:45 and the next scheduled run at 01:40, so barely six hours of arrivals sat between them. The two full days that follow are the honest sample, and they brought 16 and 11 new products.
The alert file for the same three days carries exactly one, sixteen and eleven new in market lines. That is a consistency check rather than an independent confirmation, because the same run writes both files. It does rule out the one failure we can rule out cheaply, a register that grew without anybody being told.
Of the 85 cards with a price, 44 stand at 99.99 USD or below and the sorted middle value is 99.99 USD. Nine of those 44 are 0.00, which is what the page shows for a free product. At the other end 15 cards ask 299.00 USD or more and a single one asks 990.00 USD. The most frequent price above zero is 99.00 USD on eight cards.
Our own shelf sits at the other end of that range. The 20 audited products here that carry a list price run from 149 USD to 1,699 USD, and ten of them are priced at 559 USD or less. That is not a statement about the market, it is a statement about two lists. The catalog is not a sample of anything. It is the set of products we chose to test.
Thirteen cards carried no price value for the parser to read. The register stores nothing there rather than a guess, and the alert line for those products says the price is unknown. A missing price is not a free product. The nine free ones printed 0.00 and are counted as such.
This is the same rule the audits follow. A field we could not read stays empty and gets named as empty, because the alternative is a number that looks measured and is not.
The price on a card is what the vendor asked on the day we looked. Our watch has already caught that moving inside a single week. Lizard was recorded at 699 USD on 2026-08-15 and at 199 USD on 2026-08-24, and both values sit in the price history of the Lizard audit. The arrival price of a new product deserves the same suspicion.
32 of the 98 names contain gold or XAU. The register holds 86 distinct authors, ten of whom placed more than one product in the window and one of whom placed four. A name is a marketing decision and a product count is not a track record. Neither number says anything about what these products do on a chart, and the register was never built to answer that.
The catalog carries 28 products with an MQL5 id and not one of them appears in the register of 98. That is a fact about coverage rather than a verdict. An audit here means a full tester run on our own tick history, and what that produces looks like The Gold Reaper or Scalping Robot Pro, not like a line in a listing.
Every audit we publish puts a measured result next to the version and the window it came from. Your own tester report? The browser check is free.