312 MQL5 EAs Tested: Real Ticks Cut Median PF 2.46→0.92

312 MQL5 EAs Tested at Once — 60% of the Top Group Turned Loss-Making on Real Ticks

Following our mass test of 43 popular gold EAs in July 2026, this time we widened the scope and backtested 312 paid EAs from the MQL5 Market under the same conditions. We then re-tested on 100% real ticks the 78 gold EAs that passed the first screen on the 1-minute OHLC model. Every figure is our own measurement. No official images or official performance records are reproduced.

Test Conditions (Full Disclosure)

All 312 EAs were run under the same conditions. No EA had its conditions changed.

The first-screen criteria were "profitable on OHLC, PF 1.3 or higher, 300+ trades, win rate below 90%". We excluded win rates of 90% or more because many high-win-rate EAs use averaging down and martingale to postpone large losses.

Result 1: The Big Picture for 312 EAs

The 33 unmeasurable EAs break down as follows: 16 produced no report (the target symbol was not available in our test environment: 13 BTCUSD, 2 NAS100, 1 USOIL), 9 timed out, and 8 were stopped because the tester log grew abnormally large. These could not be judged, and we did not rate them.

One in three EAs that traded lost money, and more than one in four lost almost the entire account. This is the result on 1-minute OHLC, a model that tends to be lenient with thin-margin scalping.

Result 2: On Real Ticks, Median PF Fell from 2.46 to 0.92

We re-tested the 78 gold EAs that passed the first screen on real ticks (Dukascopy, 100% quality). Of the 78, 66 finished, 11 timed out and 1 produced no report.

Comparing the 62 EAs that traded under both models, the results changed sharply.

All 78 in the first screen were profitable on OHLC. Even so, on real ticks 60% turned loss-making.

Examples Where PF Collapsed

The higher an EA's OHLC PF, the harder it tended to collapse. OHLC approximates price movement using only the open, high, low and close of each 1-minute bar, so the smaller an EA's take-profit, the more often it fills "take-profits that would not actually have been reached". We explain the mechanism in why real ticks and OHLC give opposite results.

Note: the real ticks in this test are Dukascopy data, whose spreads are wider than MetaQuotes' (in our past measurements, about $0.47 vs about $0.07 on average). Part of the collapse is due to the spread difference. For example, Vex Gold EA scored PF 3.00 on MetaQuotes real ticks (2020-2026) in the follow-up to the 43-EA test. The accurate reading is that its edge disappears on accounts with wide spreads.

Result 3: Only 4 Were Profitable in All 12 Years with Max DD of 30% or Less

Applying the first-screen criteria (profitable, PF 1.3 or higher, 300+ trades, win rate below 90%) plus max DD of 30% or less to the EAs that finished on real ticks left 10. We also checked their year-by-year results and excluded the following 6.

• Drifter EA: net profit of +$134 (11.7 years) — almost no profit
• Osloma Gold / Osloma FlipPro MT5: trades concentrated in 2025-2026, with almost none before that
• Logan MT5: only 470 trades, and −14.7% in 2016
• Smart Golden Portfolio: PF 1.31, net profit +$1,316, loss-making in 3 of 12 years
• AMG Gold: net profit +$2,096, loss-making in 4 of 11 years

The 4 that remained are:

All max DD figures are the relative equity drawdown over the same period (2015-2026).

We do not list net profit. EAs that compound their lot size grow to unrealistic amounts over 11 years of compounding (GOLD Dynamics Pro reached +$1.29 billion on paper; by the end of 2015 its lot size had already hit the 200-lot cap, a size that could not be filled on a live account). Please compare by PF, max DD and year-by-year stability. For how compounding looks in backtests, see the compounding illusion in backtests. Only Gold Albatross uses nearly fixed lots; its net profit was +$8,324 (+1.4 to 12% per year).

EAs Excluded Because Max DD Exceeded 30%

Even with a high PF, EAs whose max DD exceeded 30% were excluded.

MT5 reports show two figures: "maximal drawdown" (the decline that was largest in money, and its percentage at that point) and "relative drawdown" (the decline that was largest in percentage). For EAs whose balance grew large through compounding, the former's percentage looks small, so this article consistently uses the latter. We also explain the difference in equity vs balance drawdown.

Ghost Xau mt5 showed PF 915.46 on real ticks. Results with almost no losses are hard to imagine in live trading, so we excluded it from the evaluation.

EAs That Could Not Be Judged Due to Timeouts

The 11 EAs that did not finish on real ticks within 3 hours have no verdict (GOLD Phantom X, SomaGold, Chrysos, Fortress MT5, The Gold Reaper MT5, The Gold Phantom, Goldbot One MT5, GoldBaron XauUsd EA MT5, Gold Spark, Gold Trade Pro MT5, Nine Engine EA MT5). A timeout says nothing about whether the results are good or bad. For example, SomaGold was rated Excellent in a real-tick retest on another broker (HFM).

Limitations of This Test (Please Read)

• Tests used default settings on a single feed. Results may differ with vendor-recommended settings or other brokers.
• The real ticks are Dukascopy data with relatively wide spreads. On ECN accounts with tight spreads, EAs that lost money here may turn profitable.
• The mass test used the OHLC model. EAs that did not pass the first screen were not re-tested on real ticks.
• Results reflect the versions available at test time (obtained October 5, 2026). Later updated versions were not tested.
• Backtests do not guarantee future performance.

Summary (Facts Only)

• We tested 312 paid EAs from the MQL5 Market under identical conditions. Of the 263 that traded among the 279 measurable, 33% were loss-making and 27% had max DD of 90% or more.
• Re-testing the gold EAs that were profitable on OHLC with real ticks dropped the median PF from 2.46 to 0.92, and 38 of 62 turned loss-making.
• The 4 that stayed profitable in all 12 years with max DD of 30% or less were GOLD Dynamics Pro, Vortex Gold EA, X Prime EA and Gold Albatross.
• A high PF on OHLC is not, on its own, a basis for judgment. If you are considering a purchase, we recommend running the free demo on real ticks from your own account's feed and checking the max DD and year-by-year results.

We also publish all long-term backtest figures for our own EAs, together with the maximum equity drawdown (test conditions are stated on each EA page). See our verification methodology and the EA ranking.

Related articles: 43 popular gold EAs mass-tested on 11.5 years of real data / 43 gold EAs over 11.5 years: price and quality are unrelated / Why real ticks and OHLC give opposite results

Notes on Names and Rights

• The EA names in this article are trademarks or product names of their respective owners. They are used descriptively for comparison and review; we have no affiliation with the sellers.
• All figures in this article are our own backtest data. No official images or official performance records are reproduced.
• If you find a factual error, please contact us via Contact. We will check it and correct it promptly.