Iconic AI Signal

ICONIC AI SIGNAL v2.0
Adaptive Multi Timeframe Confluence Indicator for MetaTrader 5

Traditional technical indicators fail the moment you take them live. The reason is simple and mathematically unavoidable: they are static. Fixed formulas. No memory. No ability to recognize whether the market is currently trending cleanly or decaying inside a choppy, illiquid range. When they look perfect in a backtest, it is almost always because they have been unintentionally fitted to past noise. In live conditions, they break at exactly the points where they are needed most: local exhaustion zones, false breakouts, and regime transitions.

ICONIC AI SIGNAL v2.0 is built differently. It is a fully transparent, adaptive Multi Timeframe Confluence Engine running natively inside MetaTrader 5. At its core sit two parallel AI systems: a per regime logistic regression brain that learns which entry features predict success under each market condition, and a spiking neural network coprocessor that measures the temporal coherence of those features — how synchronously the confluence signals fire rather than just whether they fire. Both systems update from every resolved trade in real time and save their state across sessions. Nothing is fixed. Nothing is static.

Compatible with Forex majors and minors, Gold (XAUUSD), Silver (XAGUSD), Bitcoin (BTCUSD), Ethereum, and leading indices including DAX, US30, NASDAQ, and S&P500.

THE STRUCTURAL FOUNDATION: HALFTREND BACKBONE

The signal engine uses a HalfTrend baseline as its structural core. The HalfTrend line combines moving amplitude medians with an ATR based channel deviation, producing a directional trend line that flips color when the macro structure switches. Crucially, the system does not trade every flip. The HalfTrend direction is the prerequisite condition. The three entry pathways below determine precisely when and how a position is considered.

THREE ENTRY PATHWAYS

PATHWAY 1: PULLBACK TO EMA CONTINUATION (PRIMARY)

The highest statistical expectancy pathway in trending conditions. After the HalfTrend flips direction, the system arms itself and waits. Instead of chasing the breakout bar where exhaustion candles typically form, the engine waits for price to retrace into the zone around the 20-period EMA. When price enters this pullback zone AND a confirmation candle closes in the trend direction with a minimum body to range fraction, the signal fires. The system buys dips in uptrends and sells rallies in downtrends with defined structural logic, not impulse. An arm timeout limits how long the engine waits for the pullback before deactivating.

PATHWAY 2: MOMENTUM BREAKOUT

For explosive, high momentum market phases. The engine targets a genuine breakout beyond the structural high or low of the recent lookback window. This pathway is strictly gated by the ADX momentum filter: only signals accompanied by expanding directional strength are accepted. Classic bull and bear traps in weakening trends are filtered before the signal is generated.

PATHWAY 3: FRESH FLIP ENTRY

For reversal oriented entries. Activates the moment the HalfTrend line switches direction, conditioned on the confirmation candle showing significant price expansion at the flip point. This pathway is disabled by default and best suited for more aggressive tactical use.

THE THREE REGIME ARCHITECTURE

Markets behave fundamentally differently under different conditions. A feature that predicts success in a trending environment may do the opposite in a sideways chop. ICONIC AI SIGNAL v2.0 solves this by splitting the market into three distinct regimes based on ADX readings:

• REGIME 0: LOW (Range). Sideways, choppy markets. Breakout entries bleed out in this regime. The indicator can hard skip this phase entirely via InpSkipRange to preserve capital while the system learns nothing from low quality structure.
• REGIME 1: MED (Trend). Healthy, stable trending conditions. The AI brain heavily prioritizes Pathway 1 pullback signals here where their expectancy is highest.
• REGIME 2: HIGH (Strong Blowoff). High velocity impulse expansions. The brain assigns elevated weight to breakout velocity features and momentum confirmation.

Each regime commands its own completely separate weight matrix inside the logistic brain. The three matrices learn independently and never contaminate each other.

THE AI BRAIN: PER REGIME LOGISTIC REGRESSION

The primary decision engine is a per regime online logistic regression model running directly on your processor. After every resolved signal, the system calls its training function, computes the deviation between the predicted win probability and the actual outcome, and immediately updates the relevant regime's weight matrix using Stochastic Gradient Descent with a decaying learning rate. L2 regularization penalizes extreme weight values at every update, keeping the model stable and preventing it from overfitting to recent noise. Weights are hard clamped to a symmetric range to prevent runaway divergence.

The model evaluates each setup through a 6-dimensional feature vector:

• Feature 0 (Body Strength): The trigger candle's body size relative to its full high to low range. Measures momentum quality at the moment of the signal.
• Feature 1 (RSI Health): Verifies that directional momentum is in a structurally healthy zone and not already deep into overbought or oversold territory.
• Feature 2 (ADX Rising): Binary state indicating whether market velocity is actively building at the time of entry rather than decelerating.
• Feature 3 (ADX Level): The scaled absolute trend intensity of the current regime, distinguishing mild from strong directional environments.
• Feature 4 (HTF Alignment): Higher timeframe confluence check against the H1 EMA direction. Can be soft (informational) or hard (blocking) depending on configuration.
• Feature 5 (ATR Expansion): Detects whether current ATR exceeds its own moving average reference, indicating institutional grade volume influx and genuine volatility expansion.

A minimum raw confluence score gate (InpMinScore) requires at least a configurable number of these six features to confirm before the signal is even evaluated by the AI gate. This strips marginal setups before any learning cycle is triggered.

Following a configurable bootstrap warmup period (default 25 resolved trades), the AI gate assumes full control. Any new setup whose computed win probability falls below the minimum threshold is blocked: no arrow is drawn, no alert fires, no capital is risked. The AI lift tracker in the dashboard shows the exact difference between the raw win rate (all signals) and the gated win rate (only AI approved signals) so you can verify the gate's contribution live.

QUANTUM VECTOR: eSNN 2.0 TEMPORAL COHERENCE COPROCESSOR

Version 2.0 introduces a parallel neuromorphic coprocessor running alongside the logistic brain. This is not a replacement for the calibrated regression model. It adds a dimension the regression brain cannot see: time.

The logistic brain asks whether the six features are currently active. The eSNN coprocessor asks whether they are activating synchronously. A genuine high probability setup does not just check boxes sequentially — all relevant signals tend to converge within the same short temporal window. The eSNN detects this convergence and outputs a coherence score between 0 and 1 that modulates the logistic brain's p(win) estimate in log odds space. High temporal coherence amplifies an already confident signal. Low coherence softens it.

The architecture consists of 16 cortical output neurons receiving input from the same 6 features used by the logistic brain. Input encoding is temporal latency encoding: stronger features spike earlier in the 20 step simulation window. The network runs a Leaky Integrate and Fire kernel where each neuron integrates incoming spike current, leaks between steps, and fires when it crosses its dynamic threshold. The coherence readout measures how early and how broadly the output bank fires: early, dense firing indicates high temporal synchrony of the confluence.

Three biological mechanisms make the network adaptive rather than static:

• Reward Modulated STDP: Spike Timing Dependent Plasticity with reward sign modulation. When a trade wins, synaptic pathways that contributed to that signal are strengthened via long term potentiation. When a trade loses, those same pathways are depressed via long term depression. The network learns profitability, not merely input timing. Both the logistic brain and the eSNN save to separate local files and reload across terminal restarts.
• Homeostatic Threshold Regulation: Each output neuron's firing threshold decays back toward its baseline between signals. This prevents threshold creep from silencing neurons that have fired frequently and ensures the network maintains sensitivity across varying market activity levels.
• Dendritic Pruning: Synaptic connections that fall below a minimum weight threshold and remain dormant for 500 consecutive bars are severed entirely. This keeps the network lean and prevents accumulation of vestigial pathways from historical market conditions that no longer apply.

The eSNN also receives direct shock injection from the news filter: when a high impact economic event approaches, the astrocyte energy level in the network rises, temporarily increasing membrane leak and making the coherence output more conservative during chaotic market phases.

MTF EDGE SELECTOR: AUTOMATED TIMEFRAME OPTIMIZATION

The indicator continuously runs a causal walk forward replay across M5, M15, M30, and H1 in the background. For each timeframe it tracks wins, losses, and profit factors across three rolling windows: 7 days, 30 days, and 90 days. A recency weighted scoring algorithm blending these windows identifies the current champion timeframe — the one with the best verified statistical edge right now, not historically. A minimum trade sample and minimum profit factor gate prevent the system from crowning a timeframe that has not yet proven itself with enough data. When InpMTFOnlyChampion is active, signals are only passed if your active chart matches the champion. Alternatively, the champion timeframe can be monitored in the background and its live signals plotted directly onto your current chart.

INTEGRATED NEWS FILTER

The indicator connects directly to the MetaTrader 5 native economic calendar. It automatically identifies the relevant currencies for your trading pair and monitors upcoming events in a configurable lookahead window. When a high impact release approaches, the system sends a push notification to your mobile device and suspends all signal generation for a configurable window before and after the event. Both high impact and medium impact event filtering can be configured independently.

SESSION FILTER

A configurable session filter restricts signal generation to defined server time hours (default 07:00 to 21:00). An optional Friday evening block prevents entries in the final hours before the weekly close when liquidity thins and spread conditions deteriorate.

RISK AND TRADE LEVELS

Every validated signal automatically maps structural price levels onto the chart:

• Stop Loss: Placed behind the nearest structural swing high or low within the configurable lookback, padded by a volatility derived ATR buffer. Structurally anchored, not based on fixed points.
• Take Profit 1 (TP1): Placed at exactly 1.00R. This is the primary metric the AI brain is trained on. Hitting TP1 is defined as a win. It stabilizes the equity curve and provides the clean binary outcome needed for logistic regression training.
• Take Profit 2 (TP2): Placed at 2.80R. The runner target engineered to capture macro trend extensions and substantially boost the system's long run profit factor.

SIGNAL GRADE CLASSIFICATION

Every signal that passes the AI gate is immediately classified into one of three tiers based on the combined output of the logistic brain and the eSNN coherence score:

• A Grade: Highest AI confidence. Plotted as a large arrow on the chart. Mobile push alerts fire for these by default.
• B Grade: Medium confidence. Plotted as a medium arrow. Can be included in push alerts via configuration.
• C Grade: Lowest confidence above the gate threshold. Plotted as a small arrow. Push alerts suppressed by default for this grade.

DASHBOARD AND TRANSPARENCY TOOLS

The on chart panel updates on a configurable heartbeat timer independent of incoming ticks, ensuring it never freezes during quiet market hours:

• Current signal and entry pathway classification
• AI win probability from the logistic brain and eSNN coherence score side by side
• Signal grade (A, B, or C)
• Raw win rate versus gated win rate comparison — the exact performance lift the AI gate is producing, live, in front of you
• MTF Edge Selector table showing trade sample sizes and win rates across 7d, 30d, and 90d for all monitored timeframes, with the current champion highlighted
• Economic calendar preview with upcoming events, currency indicators, impact levels, and countdown timers
• Entry, SL, TP1, and TP2 levels for the current active signal

CSV DATA EXPORTER AND ALERTS

When enabled, the CSV exporter logs the complete trade history on chart close or terminal shutdown. Every record includes the full feature vector values, computed AI probability, eSNN coherence score, signal grade, entry pathway, regime classification, and final trade outcome. The output file is suitable for direct import into Excel, Python, or R for external quantitative analysis.

Push notifications route directly to the MetaTrader mobile app on iOS and Android. The InpPushOnlyStrong option restricts mobile alerts to A grade and B grade signals only, eliminating notification noise from lower confidence setups.

TECHNICAL SPECIFICATIONS

• Platform: MetaTrader 5
• Asset classes: Forex majors and minors, Gold (XAUUSD), Silver (XAGUSD), Crude Oil, Bitcoin (BTCUSD), Ethereum, Solana, XRP, DAX, US30, NASDAQ, S&P500
• AI architecture: Per regime online logistic regression (3 regimes × 7 parameters) + eSNN 2.0 temporal coherence coprocessor (16 output neurons, 6 inputs, 20 simulation steps)
• Learning mechanism: SGD with decaying learning rate and L2 regularization (logistic brain); reward modulated STDP with homeostatic regulation and dendritic pruning (eSNN)
• MTF analysis: M5, M15, M30, H1
• Brain persistence: Two separate .brain files per symbol per timeframe, reloaded automatically across terminal restarts

PRICING

• Current Price: 99 USD
• Next Tier: 149 USD
• Final Price: 249 USD

Once licenses for a tier are exhausted the price steps up automatically. Early adopters lock in their access at the lowest tier permanently.

DEVELOPER

Developed by ICONIC.FX Trading Systems. Fully linked to the verified MQL5 developer profile of mauriceprg for code quality, regular updates, and feature requests. Support via MQL5 comments and direct messages.