Our archive holds 376 recorded version entries across 24 products, and every one of them carries text the seller wrote. 185 of those entries mention a fix. 132 announce something added. Five put a number on an outcome, and not one of the five names the window, the symbol or the account it was measured on. That ratio is the subject of this piece.
Once a night the watch fetches the updates page of each product we follow. It splits the page into update blocks and keeps three things per block, the version string, the date, and the text of the entry with the markup stripped away. Nothing is rewritten and nothing is summarised. What sits in our archive is the vendor text, cut off at 2000 characters.
That cut has already bitten four entries. One Adaptive Gold Scalper entry stops at "This mode is ideal for", a Twister Pro Scalper entry stops at "This fi", and two Waka Waka entries stop mid word at "Low I" and at "reduces the numbe". Those four tails are not in our copy and nothing below rests on them.
185 of the 376 entries contain the word fix, 54 of them use the word bug, and 132 use the word added. 38 entries do both and 97 do neither. These are word matches rather than a reading of intent, so take them as a shape rather than a classification.
The shape is plain enough. 119 entries say what they have to say in 60 characters or less. One Adaptive Gold Scalper entry reads "Fix a critical bug urgently." and that is the whole entry. Waka Waka has two that read "Bug fix release". Lizard has "Expert log spam fix" and Pulse Engine has "Fixed issue with slow backtest". A product that ships repairs is not a scandal, it is what maintained software looks like. The point is what those lines leave out, which is any statement about what the repair did to a result.
20 entries carry a percent sign and the word percent appears in none of them. 15 of the 20 describe a setting rather than a result, such as the three risk levels 0.75%, 1.5% and 2.25% that Gold Atlas lists in one entry, or the default drawdown limit that Logan lowered from 30% to 15% in version 3.14. Settings we can read in the terminal ourselves.
Five entries are a different kind of statement. Logan 1.3 of 2026-06-27 introduces a reversion strategy "increasing total profits by about 30% without increasing the drawdown". Logan 1.5 of 2026-07-06 reports that "the 2026 drawdown was reduced from 33% to just 6%" at one named broker. Logan 1.8 of 2026-07-19 says "Backtests show approximately 20% higher profits, while slightly reducing overall drawdown". Logan 2.0 of 2026-07-23 claims higher profits of 20 to 30 percent at most brokers and a 50% smaller relative equity drawdown "on some brokers in real tick data". Pulse Engine adds a small account mode that "reduced the historical risk on these markets by approximately 50%".
Every one of those sentences reports a measurement somebody made. None of them names a window, a symbol, a deposit, a cost model or a tick source, and four of them come from the same product inside four weeks. We cannot reproduce any of the five and neither can you. Nine entries use a phrase such as higher profits or increases profits, and five of those nine carry no number at all. Exactly one entry in the archive uses a word like double or twice, and it warns about duplicate orders rather than promising anything.
An audit here names the version it ran, and that version usually has an entry in this archive. Our Logan audit measures 2.95, and the 2.95 entry of 2026-08-13 opens with "Added Logan's Secret (Strategy 7)". Our Lizard audit measures 1.87, whose entry of 2026-08-12 reads "Fixed news calendar filter" plus one panel addition. Our Ultimate Breakout System audit measures 7.0, whose entry of 2026-08-12 adds a range breakout entry model. Our Gold Reaper audit measures 4.6, and that entry of 2026-05-19 reads in full "Added option for much faster backtesting with real ticks".
Those four we checked by hand rather than sweeping the whole catalog. What they show is the distance between two kinds of statement. The vendor line says what was added. The audit says what came out of one run, on our tick history, at one deposit, with one cost model, over one window. Only one of the two can be repeated by somebody else.
The archive is not a diary of updates we watched arrive. 322 of the 376 entries were recorded on 2026-08-15, the day the watch first read these pages, and the version dates behind them reach back to 2021-08-23. Later arrivals can be back catalog as well. Twister Pro Scalper brought 20 entries and Zerqon brought 3 on 2026-08-22, the day those two joined the watch, and Smart Gold Hunter brought 6 on 2026-08-16 for the same reason.
Only 22 of the 376 entries carry a version date on or after 2026-08-15. Everything older is the memory of the vendor page, and the vendor owns that page. Counted by year the archive holds 6 entries for 2021, 3 for 2022, 30 for 2023, 66 for 2024, 90 for 2025 and 181 for the part of 2026 we can see. That curve says at least as much about which products we follow as about how often anybody ships.
Every audit we publish names the build, the window and the cost model behind its numbers. Your own tester report? The browser check is free.