Blog: reading EA evidence like an auditor

Why backtests lie, how to read tester reports, and what our audits measure. The method is free; the verdicts are in the catalog.

Quantum Queen MT5 review: the long history starts late

2026-09-20

Quantum Queen earned 4,597.04 USD, but its first trade came 14.7 years into the available history and its baskets reached 17 simultaneous positions.

Read →

What our software tests can establish

2026-09-18

Constructed examples check capital bases, changing position sizes and insufficient history. Their expected answers explain software behavior, with a clear limit on what they establish about an EA.

Read →

Two tests in one journal: the field that says which run an audit read

2026-09-17

All 33 audit files in this catalog carry the tester journal next to the report and pair every trade from it. In two of them the journal holds two test runs, and only the last one counts because MT5 starts its order numbers over with every test. Logan pairs 1,372 trades out of such a file at full coverage, and in 285 of them the journal names an entry that a first in first out rule would have missed.

Read →

What the changelogs claim: 376 updates and five numbers

2026-09-17

Our market watch archives the vendor updates page of every product we follow, and the copy read on 2026-08-27 holds 376 version entries across 24 products. 185 of them mention a fix and 132 use the word added. Five put a number on an outcome, and not one of the five names the window, the symbol or the deposit it was measured on.

Read →

Thirteen sellers behind twenty six audits: what the vendor line shows

2026-09-16

The 26 default audits carry 13 seller names, and six of those names carry 19 of them. Five of the six green prop fit lamps sit under one seller, and two separately sold robots from another seller land at 986 and 984 trades with 329.72 USD and 314.65 USD of audited net on the same gold history.

Read →

Thirty rewrites in twelve days: the product page is not a fixed document

2026-09-16

Our market watch hashes every product description once a day and files a dated copy of the page when the hash moves. Between 2026-08-16 and 2026-08-27, 15 of 27 watched products rewrote their description, 30 times in total, and Scalping Robot Pro alone changed on 8 separate days. One delisted product hashes to an empty input, which is what a vanished page leaves behind.

Read →

The worst day, measured twice: what a daily loss rule sees that a trade log does not

2026-09-15

Every audit answers the question what was the worst day twice, once from the trades that closed and once from the equity curve. In 19 of our 26 default runs the two agree within 15 percent, in 6 they diverge, and all 6 carry our grid flag. Quantum Queen books 0.45 USD on its worst day while its equity fell 158.15 inside a day, and Logan books 30,247.94 USD in a single day while its equity never lost more than 6,867.38 in any day.

Read →

The stop loss that made money: 38,566 exits labelled SL that closed in profit

2026-09-14

Every audit sorts its closed trades into four exit classes and counts how many of the stopped ones ended in profit. Smart Gold Impulse closes 3,672 of 4,146 stops in the black, Scalping Robot Pro exactly one of 124,466, and six audits trail their stops without earning the trailing flag because the gate asks for nine in ten trades and not for a profitable share.

Read →

The window the vendor could not see: 13 of 26 audits have one at all

2026-09-14

Every audit cuts its trades at the release date and the last update date, and only the piece behind the last update is guaranteed unfitted. 13 of the 26 default audits carry such a window at all, its median is 46 days and 47 trades, and seven sit at exactly zero because the vendor shipped an update after our test had ended. Six of the thirteen read better out of sample and six read worse, on samples as small as 2 trades.

Read →

The numbers that move with your lot: seven that travel and three that do not

2026-09-13

Every audit sorts its own metrics into sizing invariant and sizing dependent, and 22 of our 26 default audits print the split at the top of the page. Gold Snap earned 1,922.00 USD at 0.08 percent compound growth on the 100,000 USD frame, and the same file reports 3,769 trades on 2,603 of 6,075 trading days with an average holding time of 1,550 seconds. Only one of those two halves survives a change of position size.

Read →

The price of a new listing: 98 arrivals and a median of 99.99 USD

2026-09-13

Our nightly market 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, the sorted middle asks 99.99 USD, nine ask 0.00 and 13 cards carried no price we could read. The 20 audited products on this site that carry a list price start at 149 USD and run to 1,699 USD.

Read →

The lamp that cannot turn red: how our six verdict lamps are set

2026-09-12

Every audit page opens with six lamps, and across our 26 default audits those 156 lamps were set by 145 individual flags. 23 of the 26 carry at least one red and not one reads ok on all six. The regime lamp is red on none of them, because none of the three rules that can set it is allowed to go past caution.

Read →

The days it stayed out: what share of the window each robot traded

2026-09-11

Every audit counts the days on which a trade closed and sets them against the weekdays in the tested window. Prop Firm Gold EA closed something on 5,533 of 6,075 days, Logan on 381 of the same 6,075. Ten of the twenty two records that carry the field closed a trade on fewer than one weekday in four, and the busiest record in the catalog still ends below zero after costs.

Read →

The findings we can name: 145 flags across 26 audited backtests

2026-09-11

Our engine defines 32 named checks and 19 of them have ever appeared in a published audit. Across the 26 default audits it has raised 145 flags under 22 distinct names, 45 red and 59 caution and 41 info. The most common finding is a breached prop firm consistency rule in 19 audits, and 13 checks have never fired at all, mostly because our fixed test conditions make them impossible.

Read →

The busiest hour: 15 of 26 audits crowd into the same three hours of the day

2026-09-10

Every audit records the hour at which trades were opened. Across the 26 default audits the busiest hour of the largest setup group falls on 15, 16 or 17 for fifteen of them, and eleven sit on hour 15 alone. Prop Firm Gold EA is the sharpest case, with 4,759 trades in one group that all opened inside a single hour of the day.

Read →

Ten caveats in every audit: what the engine writes about its own numbers

2026-09-09

Every measured audit carries ten passages the engine writes about its own method, seven in the Monte Carlo section and three in the withdrawal replay. Four of them name the direction of their own error, and Lizard shows why that matters. Its floor breach probability is 0.0 on our 100,000 USD frame and 1.0 against the 250 USD deposit the vendor recommends, where the same trade path draws down 1,018.32 percent.

Read →

The backtest that predates the product: 20 audits split at their release date

2026-09-09

Every audit cuts its trade list at the product release date and at the date of the last update. Gold Snap closed 3,720 of its 3,769 trades before the product existed, seven of 20 audits have no out of sample window at all because the vendor shipped an update 10 to 18 days after our last trade, and in 15 of 18 the window the vendor could still watch beats the decades before it.

Read →

Run it again: what 26 audits record about repeating a backtest

2026-09-08

A backtest is one run, and whether a second run would produce the same trades is a separate question. Our 26 default audits answer it in one field. Three record that the run does not repeat, 19 that it does, and 4 built by an older engine carry no answer at all. Market Anomalies EA prints 229 randomization lines and has no field, Gold Atlas has the field and prints none.

Read →

Seven million bars or five thousand: what the tester actually read

2026-09-08

Four audits of the same gold history over the same window ran on 7,901,463 bars at M1, 559,285 at M15, 280,941 at M30 and 140,862 at H1. The coverage ratio published beside those counts reads 0.9425, 1.0007, 1.0053 and 1.0081, because every bar is credited its full length while the denominator assumes 1,380 trading minutes on each weekday. Above 1.0000 means rounding, not extra data.

Read →

One version of fifty two: what the update history says about an audit

2026-09-07

Every audit page names the version it measured, the number of recorded version entries and the price on the day we checked it. Ultimate Breakout System counts 52 entries and four further builds after the 7.0 we tested, Logan records 35 entries in nine weeks, and Smart Gold Impulse has shipped none. Four products changed price inside our archive window, one of them three times.

Read →

One run, two capitals: which audit numbers move when the account does

2026-09-06

Every catalog run is measured on a 100,000 USD frame with a fixed 0.01 lot and none of them is judged on it. The 22 audits that carry the block evaluate on vendor deposits from 100 to 2,000 USD, six of them on our 1,000 USD fallback because the vendor names none. Scalping Robot Pro survives the frame at 46,765.32 USD and kills a 1,000 USD account 53 times over the same trades.

Read →

Nine of twenty earned less than they cost: list price against measured net profit

2026-09-05

Lizard closed 21,524 trades on gold and finished 183.99 USD ahead, and its licence costs 199 USD. Twenty of our twenty six default audits carry a list price next to a measured net result, and nine of them earned less across the whole test window than the software costs to buy. One 299 USD purchase of Smart Gold Hunter contains four presets whose results lie 3,467.13 USD apart.

Read →

Five comparison files: what the catalog will not compare

2026-09-04

Our audits folder holds 38 files and five of them contain no run of their own. They compare three versions of Quantum Titan, two of Scalping Robot Pro, four presets of Smart Gold Hunter, two of The Gold Reaper, and two products that turned out to be deal identical. The part worth reading is the gate that runs first, five world fields that must agree before a single number is printed, and the note the engine writes when the test windows end on different days.

Read →

Held past Friday: what the weekend count on an audit page really counts

2026-09-04

Two numbers under every prop firm simulation say what a robot carried into the next day and across a weekend. Lizard shows 0 of 21,524 trades, Gold Atlas 2,188 overnight of which 640 are counted as spanning weekends. The weekend test also catches a Monday to Wednesday hold, so a positive count is a ceiling, and all three profiles we run permit weekend positions so no lamp judges it.

Read →

Bigger than the deposit

2026-09-03

Six current audits show drawdowns larger than the vendor deposit. Scalping Robot Pro reaches 5,323.40 percent of 1,000 USD, and Gold House shows how the same dollar loss changes across account sizes.

Read →

Years before the first trade: the test window is not the tested history

2026-09-03

A test window that opens in 2003 is not a result tested since 2003. Across the 22 default audits that carry the measurement the first trade lands between 1 and 6,086 days after the window opens. Quantum Titan reports 6.64 effective years where Gold House reports 23.28 on the same start date, and the three runs that begin trading in 2006 do so on a history our own coverage figure puts at 1.0005.

Read →

The months that paid: what a monthly sweep would have paid out

2026-09-02

A backtest reinvests every dollar it makes and a person takes money out. Our monthly withdrawal replay puts Gold House at 182 paid months out of 280 and Lizard at 3, has Logan bank $629.48 and finish $24,971.47 below its deposit, and finds Scalping Robot Pro paying nothing in 782 of 1,000 reshuffled paths.

Read →

What holding would have paid: 13 of 22 audits earn less than the price move

2026-09-02

Every audit prices what the traded instrument did over exactly the span the EA was in the market, at exactly the lot size it used. Lizard kept 183.99 USD where the price move was 4,070.66 USD, and 13 of our 22 comparable audits finished below their own benchmark. OilVector X beats crude oil by 1.52 USD across almost nine years.

Read →

Surviving a prop firm floor is only half the test

2026-09-01

Logan avoided the measured 6% floor at one sixth size but paid 0.01% per year. Scalping Robot Pro still recorded one account death at one eighth size and paid 0.00%.

Read →

The label on every trade: what an EA writes about itself

2026-09-01

Every MT5 trade carries the comment the EA wrote when it opened, and our audits add up the result behind each label. In Lizard six labels split 21,524 trades and the largest one lost 1,163.62 USD while the other five earned 1,347.61 USD. In Prop Firm Gold EA all three labels earn, and one of them dug 595.47 USD of the 683.08 USD deepest drawdown.

Read →

Scalping Robot Pro: both audited versions lose money

2026-08-31

Versions 3.0 and 4.0 both lost more than 53,000 USD after costs and retain red structure and prop-firm warnings. The unequal history windows cannot isolate a version effect.

Read →

Stacked is not a grid: read the direction, not the ticket count

2026-08-31

Several open positions do not prove a grid. Five current audits show how split tickets, adds with the move and adverse adds lead to different structure findings.

Read →

532 Quantum Titan entry attempts were refused

2026-08-30

Quantum Titan's ledger records 1,260 refused market requests: 532 entry attempts and 728 close attempts across 1,947 completed trades.

Read →

Same trades, another order: what a thousand reshuffles show

2026-08-30

Lizard earned 183.99 USD but lost money in 22.1 percent of its bootstrap paths. Gold Trade Pro earned 13,446.15 USD and lost money in none. Two current audits show how to read resampled profit and drawdown.

Read →

Drawdown recovery can take years

2026-08-29

A small drawdown can still lock up capital for years. Lizard recovered a 2.55 percent drawdown after 6,027 trading days, while Twister Pro Scalper took 3,191 days to recover a 0.11 percent drawdown. Logan never recovered its recorded 30.06 percent drawdown before the period ended.

Read →

Eighty percent in three days: reading the concentration number

2026-08-28

Every audit prints a line like 80% of profit made in 3 days. Lizard needs 3 of 4,242 active days for a net of 183.99 USD, Gold Atlas needs 54 of 4,636 days for 8,239.71 USD, and both light the same lamp. The count measures the best days against the net result, so a low number can mean a burst driven profit or a profit that was never really there.

Read →

The Gold Reaper: weighted lots left the risk warnings unchanged

2026-08-28

The Gold Reaper earned $4,041.83 with weighted lots on and $4,042.31 with them off. Both runs retain the same risk ratings and a variable lot-size history.

Read →

Range Breakout EA review: profit concentrated in 21 days

2026-08-27

Range Breakout EA earned $2,128.44 after costs, with 80% of profit concentrated in 21 days. The audit rates concentration and prop fit red and detects trade randomization.

Read →

OilVector X review: 1,192 oil trades and $1.52 more than buy and hold

2026-08-26

OilVector X is the only crude oil EA in our catalog. Over 8.9 years of WTI ticks it netted $35.23 after commission, while buying and holding the same 0.01 lots produced $33.71. Costs took 72% of gross profit and 80% of the result came from 8 trading days.

Read →

Zero spread does not erase a grid's losses

2026-08-26

Logan lost 24,334.65 USD after costs. Its commission and financing explain part of the loss, but low trading costs alone cannot establish grid survival.

Read →

TwisterPro Scalper review: frequent wins still leave losing years

2026-08-25

TwisterPro Scalper version 3.20 earned 329.72 USD after costs, but its worst year lost 72.09 USD despite a 74.1% win rate.

Read →

Ultimate Breakout System review: it loses half its trades and earns anyway

2026-08-25

Ultimate Breakout System wins 49.9 percent of its trades in our audit of version 7.0 and still earns 1,271.79 USD across 945 trades since 2003. Swap costs more than commission here, and the version we measured is no longer the one on sale.

Read →

ThunderGold Scalper review: almost eleven years underwater for 30 dollars

2026-08-24

The tester header of ThunderGold Scalper reads Sharpe 73 and a drawdown of 0.03 percent. Our end of day view gives Sharpe 1.11, and one drawdown ran from 2013 to 2024 for 30.41 USD. The entire record is 984 trades in twenty years.

Read →

Smart Gold Impulse review: profitable, but sensitive to costs

2026-08-23

Smart Gold Impulse earned 926.20 USD after modelled commission, but its high win rate masks cost pressure and an incomplete trading history.

Read →

Scalping Robot Pro review: 302,798 trades that end below the deposit

2026-08-22

Scalping Robot Pro v4.0 over 23 years of gold ticks. The tester head prints a profit factor of 0.88, and after our commission retrofit the run closes 53,234.68 USD below the 100,000 USD measurement frame. The audit-based read of a basket scalper.

Read →

Smart Gold Hunter review: 63 percent winners and a losing account

2026-08-22

The default preset of Smart Gold Hunter wins 63 percent of its trades and still ends 23 years of gold below the starting balance. We measured all four vendor presets on the same world. Two of them lose money and the best one earns 25 cents a trade.

Read →

Prop Firm Gold EA review: profit with a consistency problem

2026-08-21

Prop Firm Gold EA earned 6,769.79 USD after modeled commission, while its annual concentration check flagged 7 years. What that can tell you about payouts.

Read →

Quantum Athena X review: the 23 year backtest that starts in 2018

2026-08-21

The tester was offered gold ticks back to 2003. Quantum Athena X placed its first trade in January 2018 and left 14 years and 8 months of the window empty. A deal level read of 3,976 trades, the grid underneath them and what they earned against simply owning gold.

Read →

Logan MT5 review of one day that erased years of gains

2026-08-20

Logan lost 24,334.65 USD after costs. Its current audit flags a large basket loss and years without trades at the start of the test.

Read →

Market Anomalies EA review archived

2026-08-20

The original Market Anomalies EA review has been archived because its test evidence was withdrawn. A replacement audit is not yet available.

Read →

Lizard EA review: 21,524 trades over 23 years, and $183.99 left

2026-08-19

Lizard kept 183.99 USD after costs across 21,524 trades, while a 250 USD historical replay reached zero from 61.0% of starting points.

Read →

Gold Snap review: a narrow edge concentrated in a few days

2026-08-18

Gold Snap made 1,922 dollars after costs, but 80 percent of its profit came from 169 trading days. Its 51 cent average trade leaves little room for higher costs.

Read →

Gold Atlas review: profit concentrated in a few days

2026-08-17

Gold Atlas made $8,239.71 after costs, with 80% of profit in 54 days. An extra $1.00 per trade turns the run negative, while smaller accounts face larger drawdowns.

Read →

Gold House MT5 review: the first four years are the honest part

2026-08-17

Gold House lost money in 2003, 2004, 2005 and 2006 and has been profitable in every year since. It reached the market in February 2026 with 23 years of backtest behind it. A deal level look at four pending order strategies and an edge of one dollar per trade.

Read →

CryonX EA MT5 review: $0.84 of cost headroom per trade

2026-08-16

CryonX earned $3,456.01 in the gold backtest. An extra $1 per trade turns the result negative, while the same trade sizes carry substantial drawdown on a small account.

Read →

Adaptive Gold Scalper MT5 review: account size changes the drawdown

2026-08-15

Adaptive Gold Scalper earned $381.48 after costs. Its modeled drawdown rises from 0.24% on $100,000 to 47.62% on $500 at the same trade sizes.

Read →

Quantum Titan review: three versions in two weeks, measured

2026-08-15

Quantum Titan appeared on 2026-08-07 at 1,199.99 USD and shipped two updates within two weeks. We built and audited all three versions on the same 23 years of gold ticks. Version 3.0 doubles the backtest of 2.2 and clears the prop lamp. There is no out of sample window at all.

Read →

Gold Trade Pro earns a profit but fails the consistency check

2026-08-14

Gold Trade Pro earned $13,446.15 after costs. Its annual consistency check flags 8 years despite zero modeled loss limit breaches. What those results mean.

Read →

Waka Waka review: the previous results are retired

2026-08-14

The 2026-08-13 results are retired. A full-history measurement of version 4.59 is complete, but a reviewed replacement verdict is still pending.

Read →

Quantum Queen X review: the delisted Queen, republished and measured

2026-08-13

Quantum Queen was delisted; Quantum Queen X appeared. Our deal-level census found both tester runs identical: 6,714 deals, matching to the cent. The audit-based read of the republished grid, over 23 years of gold ticks.

Read →

The Gold Reaper review: 80% of the profit came from 31 days

2026-08-13

The strongest tester run in our catalog still concentrates 80% of its profit into 3.5% of its trading days. What that calendar problem means for anyone starting today, measured over 6.6 years and 4,258 trades.

Read →