Neuro Future GOLD

Multi-Strategy Neural Network Advisor with 350 built-in virtual strategies operating on the principle of averaging with virtual deposit distribution and an ensemble of 7 trained neural networks. Developed for XAUUSD, GOLD M15. Requires no optimization, no neural network files, everything is hardcoded inside.

Recommended parameters:

• Currency pair: XAUUSD, GOLD.
• Timeframe: M15.
• Deposit from 5000 USD with standard parameters (default settings). For smaller deposits, you need to increase LotScale or decrease AutoLotEveryBalance (e.g., for a 500 USD deposit: LotScale=10, AutoLotEveryBalance=5000 or LotScale=1, AutoLotEveryBalance=500). In this case, potential profit relative to deposit will increase 10 times, but drawdown will also increase 10 times.
• It is recommended to use USD deposits, but if you use another currency, with AutoLot=true you will need to adjust the AutoLotEveryBalance parameter to the exchange rate against the dollar.
• Adjust the ManualCommission parameter, default =15.

Unique neural network architecture:

• Based on 7 deep neural networks developed using the neural network indicator Neuro Future.
• Training: Neural networks were trained on historical data for 3-3.5 years (up to July 2026) on the H4 timeframe. Forecast is made for 6 bars ahead (24 hours).
• Diversity of analysis: Each of the 7 networks analyzes the market in a unique way, using different methods and versions of the neuro indicator.
• All neural networks are already "hardcoded" into the advisor's code along with the indicator. You do not need to copy files to terminal folders.

Ensemble of 350 virtual strategies:

• Inside the advisor, 350 virtual strategies continuously operate, each with its own virtual balance, virtual positions, and parameters.
• Daily (by default) balances are redistributed (distribution) among all virtual strategies, which distributes the load and increases system stability.
• Each strategy uses a signal from one of the 7 neural networks. A signal to open is considered confirmed when the neural network outputs one direction with three buffers exceeding certain minimum thresholds.
• Ensemble optimization: Out of 2100 generated random strategies, the best 350 were selected, showing maximum performance in terms of profit/drawdown ratio, ensuring stability on unseen periods. A double selection and training system was applied. First, neural networks were trained up to 2025, testing was conducted on 2026 to identify optimal distribution parameters and virtual parameters for each strategy. Then, neural networks were retrained on new periods up to July 2026, while internal virtual parameters were left optimally from forward testing on 2025 neural networks.
• You do not need to perform additional optimizations—everything is already assembled optimally.
• The actual trading lot is formed based on the sum of all active virtual positions, lot sizing parameters, and deposit size.

Trading logic and position management:

• Grid averaging: Uses non-aggressive increasing lot coefficients to manage a series of positions. You can try using higher timeframes; grid step increases, trade frequency decreases, and vice versa. But the advisor was developed on the M15 timeframe.
• Virtual parameters: All parameters of virtual strategies (step, profit, coefficients) dynamically depend on the pair's volatility on the current period.
• Signal reversal: When the neural network changes direction, the virtual position is closed and reopened in the opposite direction with the total lot.

Filters and protection:

• News filter: Ability to close positions before important news and resume trading after their release.
• Time filter: Setting allowed trading hours.
• Auto-restart and distribution: Built-in mechanisms for automatic balancing and redistribution of virtual balances and positions between strategies.
• To preserve virtual positions, as well as temporary neural network files, the advisor uses files. If an interruption or terminal restart occurs, the advisor will load saved positions from these files and continue working (works on computers and remote systems with a desktop).

Advisor parameters:

===== MAIN PARAMETERS =====

• AutoLot ( true ) - Enable automatic lot sizing. If false, lot sizing will depend only on the LotScale parameter.
• AutoLotEveryBalance ( 5000 ) - Balance for determining autolot. Lot ~ Balance / AutoLotEveryBalance * LotScale.
• LotScale (<1 - Less Risks, >1 - More Risks) ( 1 ) - Lot scaling.
• Max.Total Lots, relative value ( 10 ) - Maximum total lot volume, relative value.
• AutoFixedBalance ( true ) - Automatic balance fixation.
• Fixed Balance (if=0 then FixedBalance=AccountBalance) ( 0 ) - Fixed balance.
• Slippage ( 20 ) - Slippage.
• SpreadLimit ( 50 ) - Spread limit.
• ManualCommission ( 15 ) - Commission.
• Magic Number ( 421556 ) - Magic number.
• FIFO ( false ) - FIFO mode.

===== V-STRATEGIES PARAMETERS =====

• VirtualMaxLots ( 5 ) - Virtual maximum lot volume.
• VirtualMaxOrders ( 10 ) - Virtual maximum orders.

===== AUTO RESTART PARAMETERS =====

• Auto Restart ( true ) - Auto restart.
• Restart Percent ( 0.5 ) - Restart percentage.

===== DISTRIBUTION =====

• Distribution ( true ) - Distribution.
• DistrTimeFrame ( PERIOD_D1 ) - Timeframe for distribution.
• NoDistribWithoutLoss ( true ) - No distribution without loss.
• HourDistributeFrom ( 4 ) - Distribution start hour.
• Smoothness ( 1 ) - Smoothness. For example, if =2, half is distributed, and the other half remains in the basket for the next distribution (displayed on the info panel).

===== Time Filter =====

• Time Filter (Set On Greenwich) ( false ) - Time filter.
• HourStart ( 18 ) - Start hour.
• HourEnd ( 13 ) - End hour.

===== SAFE CLOSING PARAMETERS =====

• Close and Remember Before Weekend ( true ) - Close and remember before weekend.
• Close All And Restart After DD,% ( 100 ) - Close all and restart after drawdown.
• Close And Remember After DD,% ( 100 ) - Close and remember after drawdown.
• Stop Days After DD (if 0-no stop) ( 2 ) - Stop days after drawdown.
• MarginControl ( true ) - Enable margin control.
• MarginCriticalLevel (100) - Critical level at which the advisor will close positions to exceed RestoreMarginLevel.
• RestoreMarginLevel (120) - Margin level for restoration.

===== NEWS FILTER =====

• NewsFilter ( false ) - News filter.
• HighImpact ( true ) - High impact news.
• MiddleImpact ( false ) - Medium impact news.
• LowImpact ( false ) - Low impact news.
• Speaks ( true ) - Speaks.
• USDNewsForAll ( true ) - USD news for all pairs.
• MinsBeforeNews ( 5 ) - Minutes before news.
• MinsAfterNews ( 30 ) - Minutes after news.
• Show alerts ( true ) - Show alerts.
• Send notifications ( false ) - Send notifications.