Iconic Neurocore AI

ICONIC NEUROCORE AI+  |  OMNI NEXUS EDITION
Dual Symbol Expert Advisor for BTCUSD and XAUUSD on MetaTrader 5  |  Version 2.00

Running two unrelated automated systems on the same account is not the same as running a coordinated two asset architecture. Two independent EAs share margin and equity but cannot see each other. They cannot reduce exposure when the other is already positioned. They cannot recognize when both instruments are temporarily correlated and avoid doubling a bet. They cannot determine which asset is currently leading the other and use that information to improve timing on the lagging one.

ICONIC NEUROCORE AI+ was built to close that gap. It places two specialist trading engines for BTCUSD and XAUUSD under a shared coordination layer that monitors the live relationship between both markets, allocates capital through risk parity, enforces a combined margin ceiling, and extracts a joint market regime through a sparse in RAM reservoir network. Each engine carries its own adaptive AI decision brain, its own market logic, and its own risk parameters. The coordination layer sits above both and adjusts their combined behaviour in real time without overriding their individual judgements.

ARCHITECTURE

The system operates on two levels. The first level consists of two independent symbol engines, one for BTCUSD and one for XAUUSD, each with its own AI brain, feature set, market filters, entry logic, and risk control. The second level is the OMNI NEXUS CENTRALIZED COMPILER, a coordination kernel that reads the current state of both engines together and applies portfolio level adjustments to sizing, risk budgets, and participation decisions.

Bitcoin and Gold behave differently and the architecture treats them as such. The BTC engine is built around breakout structures and rapid directional moves. It uses eight features to evaluate entries: trend state, trend strength, volatility regime, session context, day bias, drawdown fraction, ATR expansion, and spread quality. The Gold engine operates with twenty features, adding ten interaction terms on top of its core eight, and extends entry validation with H4 higher timeframe trend alignment, liquidity sweep detection on structural levels, compressed market state classification, and strong breakout candle confirmation. Gold also carries a dedicated loss streak cooldown that can block the engine entirely when a configurable consecutive loss count is reached — a control that the BTC engine does not need given its different volatility and response profile.

PER SYMBOL AI DECISION ENGINES

Each symbol engine contains an independent adaptive AI brain. The brain maintains a Q table with 36 state slots, indexed by three live market dimensions: whether the current price is in a trend, which of three volatility regimes the ATR currently reflects, and which trading session is active. These three dimensions map to a compact state representation that the brain uses to accumulate learned values for each of its five execution stances.

Action selection blends two models. The first is the Q table itself, whose accumulated visit normalized values represent what the engine has learned about each stance under each market state. The second is a linear feature weight model that scores each stance against the current feature vector directly. When an engine is new to a particular state, the linear model dominates the selection. As visits accumulate, the Q table takes over proportionally. Final action selection runs through Boltzmann temperature softmax over the blended scores, which means the engine retains calibrated exploration rather than always picking the highest scoring action. During loss streaks the AGGRESSIVE stance is suppressed and the CONFIDENT stance is progressively damped.

After each closed trade, the brain updates both components. The Q table is updated through eligibility traces: a backward looking record of which state action pairs were responsible for the current outcome. Every state action pair with a nonzero trace receives a proportional Q value update scaled by the reward. The feature weight matrix is updated through a direct reward weighted gradient on the features that were active at the time of the decision. The reward signal itself is the trade profit normalized to the risk taken, adjusted with a duration bonus for fast profitable captures and a penalty for adverse excursion during the trade lifetime. This reward structure encourages the engine to favour clean entries that produce profit quickly without large drawdowns on the position.

Both engines save their Q tables, visit counts, feature weights, and win rate history to separate local binary files and reload them after terminal restarts. Learning accumulates across sessions.

FIVE EXECUTION STANCES PER SYMBOL

The AI brain selects one of five execution stances for every evaluated setup. The stance controls lot size, stop loss placement, and take profit parameters within configured limits:

• SKIP: No setup is placed. The brain or quality gate has rejected current conditions. The OMNI NEXUS portfolio layer can also force SKIP through its protection gates.
• DEFENSIVE: Reduced lot size. Active when AI confidence is below the adaptive threshold, when Tier 1 portfolio drawdown is active, or when the OMNI NEXUS stress score is elevated.
• NEUTRAL: Standard execution using configured base parameters.
• CONFIDENT: Increased lot size within configured limits. Requires high AI confidence and neutral or low portfolio stress.
• AGGRESSIVE: Maximum configured lot size. Suppressed during any active portfolio protection tier and during loss streaks on the requesting engine.

Before the stance is finalised, the OMNI NEXUS stress score is applied as a confidence penalty. Under high portfolio stress both engines are pushed toward DEFENSIVE even if their individual AI confidence would otherwise permit higher stances. This means individual engine intelligence and portfolio coordination reinforce each other rather than operating in conflict.

THE OMNI NEXUS CENTRALIZED COMPILER

The OMNI NEXUS layer runs four interlocking technologies. Each one addresses a structural limitation of operating two independent engines on a shared account without coordination.

DIRECTED INFORMATION FLUX VIA TRANSFER ENTROPY

Standard correlation is symmetric. It measures whether two assets move together but says nothing about which one is leading and which is reacting. Transfer entropy is asymmetric. It quantifies how much the recent history of one asset reduces uncertainty about the next step of the other, computed separately in each direction. ICONIC NEUROCORE AI+ computes Transfer Entropy from BTC to Gold and from Gold to BTC across recent H1 log returns using a rank bin frequency estimator. The difference between both values produces the net information flux: a positive value means BTC is currently the informational leader, a negative value means Gold leads. When one asset is clearly leading and already holds an open position in a given direction, the lagging engine receives a bounded sizing bonus for entries in the same direction. This is a causal signal derived from the actual recent information structure between the two markets, not a static assumption.

IN RAM ECHO STATE RESERVOIR

The echo state reservoir is a sparse recurrent neural network held entirely in RAM. It receives eight inputs on each update: BTC H1 log return, Gold H1 log return, their absolute values, the current rolling correlation, the BTC to Gold transfer entropy value, the Gold to BTC transfer entropy value, and the combined margin utilization ratio. The reservoir state is advanced through a fixed sparse recurrent weight matrix and the output readout is updated online through incremental ridge regression. The result is a regime score between minus one and plus one that reflects whether the joint BTC and Gold environment is trending, compressing, or transitioning. This score modulates lot sizing across both engines: high volatility regimes reduce sizing, calm regimes allow a modest positive adjustment.

INVERSE VOLATILITY RISK PARITY

The OMNI NEXUS compiler tracks the rolling volatility of BTC and Gold H1 log returns alongside their covariance. Capital allocation between the two engines follows inverse volatility weighting: the lower volatility asset receives the larger share of the configured global risk budget, targeting equal risk contribution from each engine rather than equal lot exposure. A 30 percent floor per engine ensures that even when BTC volatility is substantially higher than Gold, the BTC engine always receives enough budget for viable lot sizing. When both engines are simultaneously positioned in the same direction and rolling correlation is elevated, concurrent positions receive proportional sizing reductions to prevent the system from inadvertently stacking correlated exposure. When positions are in opposite directions under positive correlation, a small diversification bonus applies instead.

MARGIN UTILIZATION GOVERNOR

Regardless of how tightly each engine controls its individual trade sizing, simultaneous positions on BTCUSD and XAUUSD can produce a combined margin footprint that approaches broker limits under volatile conditions. The margin utilization governor computes the ratio of used margin to total available margin across the entire account on every update cycle. When this ratio exceeds the configured ceiling, a global scaling factor reduces all new lot calculations proportionally until margin usage drops back below the limit. This backstop operates independently of all other risk controls and cannot be overridden by the per engine AI decisions.

THREE TIER PORTFOLIO PROTECTION

The OMNI NEXUS compiler tracks combined intraday drawdown across both engines against the equity at session start. As drawdown crosses configured thresholds, protection escalates:

• TIER 1: Forces both engines to DEFENSIVE sizing. Any pending CONFIDENT or AGGRESSIVE classification is overridden. Existing open positions continue to be managed.
• TIER 2: Blocks all new pending orders across both engines. No additional exposure is taken until drawdown recovers below the Tier 2 threshold.
• TIER 3: Emergency exit. All open positions across both BTCUSD and XAUUSD are closed immediately. All pending orders are deleted. A configurable cooldown period then prevents any new setups. The emergency exit state resets automatically at midnight.

A cross engine consecutive loss counter runs independently of the drawdown tiers. When combined losses from both engines exceed the configured streak threshold, a global cooldown activates regardless of the current drawdown level. Any win from either engine resets the combined streak counter.

MARKET CONTEXT AND ENTRY LOGIC

Both engines use structured pending order placement rather than immediate market execution. Per symbol validation runs first: EMA trend direction and slope, ATR volatility state, spread filter, current and previous day high and low levels, support and resistance areas from price structure lookback, order block detection, session context, cooldown status, and news filter status. The OMNI NEXUS portfolio gates are evaluated after per symbol validation and can block entry regardless of per symbol conditions: emergency exit status, global cooldown, Tier 2 drawdown block, and concurrent position capacity limits.

Gold specific validation extends this with H4 higher timeframe trend confirmation, liquidity sweep classification on structural levels, compressed market state detection, strong breakout candle confirmation, and a configurable dedicated loss streak cooldown that applies only to the Gold engine. The Gold engine also deletes stale pending orders that have exceeded a configurable maximum age, rather than refreshing them indefinitely.

RISK CONTROL

• Separate percentage based lot sizing per symbol calibrated to account equity
• Separate drawdown sensitive lot reduction per engine
• Global portfolio risk cap as dollar ceiling across both engines
• Inverse volatility risk parity split of the global cap with 30 percent per engine floor
• Correlation aware same direction sizing reduction during elevated correlation regimes
• Margin utilization governor as hard margin ceiling across both engines combined
• Adaptive cooldown multiplier per engine escalating with consecutive loss count
• Cross engine combined loss streak global cooldown
• Separate ATR based stop loss and take profit parameters per symbol
• Separate configurable risk to reward ratios per symbol
• Daily trade count limits per engine
• Global daily profit target that stops both engines when reached
• Recovery mode per engine with configurable multiplier and trade count
• Gold specific dedicated loss streak cooldown independent of the cross engine counter

TRADE MANAGEMENT

Once a position is open on either symbol, that engine manages it independently according to configured rules without requiring manual intervention:

• Break even activation after configurable favourable price movement in R multiples
• Trailing stop adjustment as price extends beyond the activation threshold
• Automatic pending order cleanup on position entry or closure
• MAE tracking across the position lifetime for AI reward calculation on close
• Daily trade count enforcement per engine
• Global daily profit target latch across both engines combined

NEWS FILTER

A single shared news filter covers both engines. It is driven by the MetaTrader 5 economic calendar and loads a configurable forward window of upcoming economic events. When an event falls within the configured time buffer, new pending orders are paused across both engines simultaneously. High impact and medium impact event filtering can be configured separately. No external data feed or independent service is required. WebRequest must be enabled in the MetaTrader 5 terminal settings for the filter to function correctly.

ON CHART DASHBOARD

The live interface displays account balance, equity, daily profit and loss, BTC engine AI action and confidence score, Gold engine AI action and confidence score, rolling win rates per engine, active execution stances per engine, trend and volatility status per symbol, spread per symbol, portfolio drawdown level, OMNI NEXUS stress score, margin utilization, net transfer entropy direction, reservoir regime score, and the next scheduled news event when the filter is active. The dashboard uses a purple and black colour theme. BTC engine elements display in cyan and purple. Gold engine elements display in yellow and orange.

SETUP AND DEPLOYMENT

• Symbols: BTCUSD and XAUUSD
• Timeframe: M10
• Platform: MetaTrader 5
• Account type: Hedging recommended
• Broker: Low spread or ECN/RAW recommended
• VPS: Strongly recommended for 24/7 stable operation

Market_Validation_Mode must be set to FALSE before live or demo operation. This parameter exists exclusively for MQL5 marketplace validation. The EA will not trade normally with it enabled.

Before deploying on a live account, verify BTCUSD and XAUUSD symbol specifications at your broker separately. Symbol names, tick values, contract sizes, minimum lots, leverage, margin requirements, and average spreads differ between brokers and will affect lot sizing and risk calculation for both engines. If the news filter is enabled, configure WebRequest permissions in the terminal settings before going live.

SIGNAL MONITORING

Live performance monitoring is available via the seller's MQL5 signal profile. After purchase, contact the seller through the MQL5 messaging system to receive setup documentation and product update announcements. Support is provided via MQL5 comments and direct messages. When reporting issues, include the broker name, symbol names, account type, terminal version, journal log output, and the exact behaviour observed. This allows setup issues to be identified faster and more precisely.

RISK DISCLOSURE

Trading BTCUSD, XAUUSD, and other leveraged instruments involves substantial risk. Automated trading does not eliminate market risk. Past performance in live monitoring, forward testing, or backtesting does not guarantee future results. Actual performance depends on broker conditions, spread, slippage, execution quality, leverage, account size, margin requirements, VPS stability, and the specific settings configured by the user. Only trade with capital you can afford to lose. Always begin with a demo account before deploying on a live account.