FXEA365 Gold Neuron EA MT4

GOLD NEURON v3.17 is a trading system for XAUUSD built around four independent decision engines that run together in a single account.

MT4 EDITION — This is the MetaTrader 4 build of GOLD NEURON (v1.00), ported from the MetaTrader 5 edition with the same engines, the same inputs and the same default values. MetaTrader 5 edition: https://www.mql5.com/en/market/product/187329 — Unless a figure is marked MT4, every figure in this description was measured on MetaTrader 5. The MT4 build has not been reconciled trade by trade against the MT5 build, and MT4 results will differ because tick model, spread and execution differ between the two platforms and between brokers.

BACKTEST OF THIS MT4 EDITION (v1.00) — MetaTrader 4 Strategy Tester, XAUUSD (broker symbol GOLD), M1 chart, every tick (modelling quality 25%), 2025.01.01-2026.07.01, deposit 10,000 USD, default inputs
- +37,185 USD / PF 1.96 / 775 trades / maximum drawdown 20.64% (10,643 USD)
This is an 18-month window in a rising gold market, and the MT4 tester builds its ticks from minute bars, so it is far more favourable than the 22.5-year MetaTrader 5 real-tick test quoted below (Profit Factor 1.35, maximum drawdown 28.50%). Size the account by the long-test drawdown, not by this window.

=== NEW IN VERSION 3.20: A FIFTH ENGINE, VORTEX ===
Version 3.20 adds a fifth engine, VORTEX, on top of the four engines described below; the rest of this description still refers to the four-engine versions. VORTEX reads the H1 CCI and opens one position with a stop loss of 33 and a take profit of 10 in price units, then places up to five limit orders of the same lot size every 1.75 against the entry, all with the same stop loss level; the whole group is closed when its combined floating profit reaches the basket target. It never increases the lot size, but it does add positions against the market (averaging down). Its lot is the balance x 0.0000025 (0.025 lot per 10,000) times the G_LotMult input (default 1.0); it is sized on top of the other engines, uses its own magic number 20261107, trades only on gold symbols and can be switched off with UseEngineVortex. Its entry filter uses lookup tables built from historical data that overlap the test windows below, so these figures are in-sample. Comparison under identical conditions (Exness XAUUSDm, M30, one-minute OHLC modelling, deposit 10,000, default settings): January 2024 to June 2026, version 3.19 net profit 25,052, profit factor 1.57, equity drawdown 24.57% -> version 3.20 net profit 91,264, profit factor 1.86, equity drawdown 19.12%; January 2015 to December 2023, 48,427 / 1.28 / 25.94% -> 702,836 / 1.73 / 21.88%. With G_LotMult 0.5: 47,192 / 1.72 / 21.44% and 161,226 / 1.45 / 22.97%. These are backtest results, not live trading, and they do not indicate future performance.

AT A GLANCE (MT5 backtest — historical, not a live record)
- Four engines in one EA on XAUUSD: an ATR-breakout trend core, a momentum breakout, a gradient-boosting entry gate and the BLITZ M15 Donchian breakout
- v3.16 as shipped (Ultra, 2022.01-2026.08, XM one-minute OHLC data): Profit Factor 1.19 / relative equity drawdown 36.85% / 595 trades
- Long window, all four engines (2004.01-2026.06, 22.5 years, real ticks): Ultra Profit Factor 1.35 / maximum equity drawdown 28.50% (largest relative drawdown 55.54%); Standard Profit Factor 1.34 / 19.95% (largest relative 39.01%)
- Win rate about 53% in the long test; averaging groups are closed by the EA's group stop. Hedging account required
- Forward test (Exness demo account, not real money; running since 2026-07-17, v3.17 since 2026-09-30, earlier versions before that): 46 trades, -4.03%, Profit Factor 0.94, max drawdown 39.38% (as of 2026-10-28). That drawdown is 1.38 times the 28.50% long-backtest maximum above; it occurred between 2026-07-28 and 2026-08-11 under an earlier v2 engine that the current version no longer uses.

=== START HERE ===
1) Read the RISK WARNING below before anything else.
2) Attach the EA to a XAUUSD chart. Any chart timeframe may be used — each engine reads its own fixed internal timeframe.
3) Test the free demo in the Strategy Tester first: see “5-MINUTE DEMO CHECKLIST” below.
4) Independent forward results: see “INDEPENDENT VERIFICATION (FXEA365)” below.
5) Platform: MetaTrader 4 (this page). MetaTrader 5 edition: https://www.mql5.com/en/market/product/187329
6) Our other three flagship EAs (FXEA365):
- Gold King Breakout Grid (XAUUSD, high-risk grid): https://www.mql5.com/en/market/product/182525
- Gold Viper Breakout (XAUUSD): https://www.mql5.com/en/market/product/182520
- AIFW AI Framework (rules engine + optional AI): https://www.mql5.com/en/market/product/196639

The decision engine was rebuilt in version 3.10, and updates are free for existing users. The four engines are a trend core based on an ATR breakout with a regime filter, a momentum breakout engine, an entry gate built with gradient boosting, and, since version 3.15, the BLITZ M15 Donchian breakout engine. BLITZ is always on and has no on/off switch; the other three engines can be switched off with UseEngineTrend, UseEngineViper and UseEnginePivot. The risk budget is split equally across the three core engines that are enabled. Since version 3.16 BLITZ is sized on its own share, one quarter of the balance, and no longer reduces the share of the other three, so the total exposure is higher than in version 3.15. The 22.5-year figures further below were measured with all four engines running. The system never increases the lot size (no martingale), but two engines do add positions against the market (averaging down): the trend core adds positions of the same lot size up to three in total and the momentum engine up to five in total, each time price moves one more ATR against the group. These two engines place no stop loss on the broker server; the EA closes each group when its combined loss equals a 1.5 ATR move on the group's total volume, checked on every tick, so the EA must be kept running. The pivot and BLITZ engines send a fixed stop loss to the broker server with every order. Logging is controlled by a single input named DetailedLog.

What changed in 3.16 and 3.17

Version 3.17 only stops drawing the on-chart panel while the EA runs in the Strategy Tester, which makes long tests much faster. Its trading logic is identical to version 3.16: in our check on 2004 to 2006 the orders and deals of the two builds matched line for line. Version 3.15 added a fourth engine, BLITZ: an M15 Donchian breakout filtered by the H1 50-period moving average and ADX, with a fixed stop loss of 6 and take profit of 30 in price units, at most three entries per day and one position at a time, under its own magic number. Version 3.16 changes how BLITZ is sized and when it may trade. First, the three core engines again receive one third of the risk budget each, as in version 3.14, and BLITZ is sized on one quarter of the balance on top of that, so the total exposure is higher than in version 3.15. Second, BLITZ opens no new position while the account balance, net of deposits and withdrawals, stands at its highest level so far; it trades only while the account is below its previous balance peak. This second rule was chosen after studying the test results, so it is fitted to the test period and may not hold in the future. The entry and exit rules of all four engines are otherwise unchanged. We compared versions 3.14 (three engines), 3.15 and 3.16 under identical conditions (XAUUSD M30, MetaQuotes-Demo data, every tick based on real ticks, where real ticks are available from May 2025 and earlier months are generated from one-minute bars, Ultra mode, deposit 10,000, leverage 1:500, January 2022 to August 2026). Net profit was 355,661 / 329,133 / 415,864, the profit factor 1.88 / 1.92 / 1.88, the recovery factor 7.54 / 9.15 / 7.71, the relative equity drawdown 57.11 / 67.86 / 54.21 percent, the maximal equity drawdown 47,146 / 35,969 / 53,919 in account currency, and the number of trades 520 / 627 / 582. Version 3.16 is not better on every figure: its recovery factor and its maximal drawdown in money are worse than those of version 3.15. A correction: the earlier comparison of versions 3.14 and 3.15 on the XM broker's data (net profit 6,653 to 8,129, profit factor 1.19 to 1.22, maximum equity drawdown 39.79 to 27.68 percent, trades 496 to 611) was published as real ticks, but that data was one-minute OHLC modelling. On the same XM one-minute data, version 3.15 versus 3.16: net profit 8,129 to 9,191, profit factor 1.22 to 1.19, recovery factor 2.83 to 2.10, relative equity drawdown 27.68 to 36.85 percent, trades 611 to 595, that is more profit with worse risk figures. A real-tick comparison on XM data will be added once measured. These are backtest results, not live trading, and they do not indicate future performance.

Backtest results

The figures below were produced in the strategy tester on XAUUSD with 100 percent real ticks, an initial deposit of 10,000 and leverage of 1:500. They are backtest results and not the results of live trading, and they do not indicate future performance.

Test window from January 2004 to June 2026, which is 22.5 years, measured with all four engines enabled (trend core, momentum breakout, entry gate and BLITZ) on Dukascopy tick data imported as a custom symbol, M30 chart. The test build is version 3.16 with panel drawing switched off in the tester, which is what version 3.17 ships. In Ultra mode the net profit was 416,042 on a deposit of 10,000, which takes the balance to 42.6 times the deposit and is an annualized return of 18.1 percent, with a profit factor of 1.35, a maximum equity drawdown of 28.50 percent (the largest relative equity drawdown was 55.54 percent), a win rate of 53.29 percent and a Sharpe ratio of 1.24 over 2,766 trades. In Standard mode the net profit was 112,152, which takes the balance to 12.2 times the deposit and is an annualized return of 11.8 percent, with a profit factor of 1.34, a maximum equity drawdown of 19.95 percent (largest relative 39.01 percent) and a Sharpe ratio of 1.36 over 2,741 trades. These figures replace the two-engine figures published before (35.2 times in Ultra and 24.1 times in Standard); the Standard result is lower than the earlier one.

This window contains the 2008 financial crisis and the gold bear market from 2011 to 2015. The win rate over the window was 53.29 percent, and long losing streaks are still to be expected in ranging markets.

Verification method

Each engine, when run standalone, reproduces the corresponding published product exactly, which confirms that the integration did not alter the logic. A forward test account has been running since August 11, 2026. Second feed verification and Monte Carlo resampling have not yet been rerun on the four-engine configuration, so no figures are published for them.

Note on the neural model used in version 2.x

The ONNX model shipped in version 2.x was trained on data from January 2020 to July 2026. Its backtest figures inside that training window are a demonstration of in-sample fit and were never obtained in live trading. In out-of-sample years from 2016 to 2019 the model was loss making in our tests, and the forward account showed negative expectancy. This is the reason the engine was replaced in version 3.10. The earlier backtest remains published for reference and is labeled as in-sample.

Risk modes

The RunMode input offers Defensive, Standard, Aggressive and Ultra. The trading signals are identical in every mode and only the position size changes. The figures above were measured in Ultra, which is the shipping default, and in Standard. Defensive halves the risk of each engine. Aggressive uses the same lot multiplier as Ultra on the trend core and a smaller one on the entry gate and BLITZ (2 instead of 3.5). Over the same 22.5-year window and conditions, Aggressive returned a net profit of 332,355 (balance 34.2 times the deposit, 17.0 percent per year) with a profit factor of 1.38 and a maximum equity drawdown of 28.00 percent (largest relative 55.18 percent) over 2,763 trades. Defensive returned 28,025 (balance 3.8 times the deposit, 6.1 percent per year) with a profit factor of 1.32 and a maximum equity drawdown of 9.96 percent (largest relative 21.40 percent) over 2,744 trades.

Recommended setup

The symbol is XAUUSD. Any chart timeframe may be used because each engine reads a fixed internal timeframe. The suggested minimum deposit is 3,000 for Ultra and 1,000 for Standard in account currency. An ECN type account is recommended and a hedging account is required.

Support

Please use the comments section of this product or the site messaging system for setup questions and update notifications.

=== WHO THIS IS FOR / NOT FOR ===
Right fit: traders who want four independent XAUUSD decision engines in one EA so that no single logic decides every trade, who value a 22-year real-tick test window over a short flashy curve, who will keep RunMode at Standard or below, and who will run a demo for at least one month before going live.
Not a fit: anyone expecting a "neural network that predicts gold" (the engines are rule-based and the gradient-boosting gate only filters entries), anyone who wants to tune dozens of parameters, anyone who needs trades every day, and anyone who cannot accept a multi-month flat or losing period inside a long track record.

=== 5-MINUTE DEMO CHECKLIST ===
1) Download the free demo and open the Strategy Tester. Symbol XAUUSD, any chart timeframe.
2) Modelling: "Every tick based on real ticks" (MT5) or 99% tick data (MT4). "Open prices only" and "1 minute OHLC" produce meaningless results for this EA.
3) Period: at least 2019.01 to today (the published window starts 2004.01). A two-week test says nothing about a basket strategy.
4) Leave the inputs at default and pick the risk preset you would actually run. Use the same deposit and leverage as your real account.
5) Compare trade count, profit factor and maximum drawdown with the figures above. A large gap usually means a short period, bad tick history, a different broker timezone, or a wrong chart.
6) Then run it on a demo account for at least one month before any live money.
After that month, please leave a review, positive or negative, with the period, broker and preset you used. Honest data helps every other trader here more than any marketing.

=== OTHER EAs WE BUILD (FXEA365) ===
Same discipline, different markets: every strategy is measured on 100% real tick data over the longest history the market allows, and every published figure comes from a test we ran ourselves. Risk profiles differ — check each page.
- Gold King Breakout Grid (XAUUSD): https://www.mql5.com/en/market/product/182525
- Gold Viper Breakout (XAUUSD): https://www.mql5.com/en/market/product/182520
- AIFW AI Framework (AI framework): https://www.mql5.com/en/market/product/196639

=== FREE TOOLS FROM US (FXEA365) ===
These are free, work with any EA including the one on this page, and are the same risk plumbing we run ourselves.
- Falcon Trailing Stop Manager: https://www.mql5.com/en/market/product/188540
- Aegis Account Protector: https://www.mql5.com/en/market/product/188539
- Spread Guard Live Spread: https://www.mql5.com/en/market/product/185749

=== A NOTE ON REVIEWS ===
If this product has been useful in your trading, a short review would help other traders and guide future improvements.