AI Adaptive Signal Trader

AI Adaptive Signal Trader

AI Adaptive Signal Trader is a fully automated Expert Advisor that combines three independent technical signals with an online, self-adjusting weighting mechanism inspired by adaptive machine-learning (multiplicative-weights / "Hedge" style) algorithms. After every closed trade, the EA rewards the signals that agreed with a winning trade and penalizes the signals that agreed with a losing trade, gradually shifting its own internal weights toward whichever signal has been more reliable in current market conditions.

How It Trades

Three independent signal modules each vote -1 (bearish), 0 (neutral), or +1 (bullish):

• Trend signal - EMA(fast) vs EMA(slow) crossover state
• Momentum signal - RSI position relative to configurable bullish/bearish levels
• Breakout signal - Price position relative to Bollinger Bands

The three votes are combined into a single weighted composite score. A trade is opened only when the composite score crosses a configurable threshold in either direction. Every position uses ATR-based Stop Loss and Take Profit, and an optional ATR-based trailing stop with built-in minimum-distance and take-profit-proximity safety checks.

Adaptive Weight Learning

When a position closes, the EA looks back at which of the three signals agreed with the trade direction and updates each signal's internal weight using a multiplicative update rule: weights of signals that agreed with a winning trade are increased, weights of signals that agreed with a losing trade are decreased. Weights are clamped to a configurable range and renormalized after every update, then saved persistently (per symbol and magic number) so learning carries over between terminal restarts.

Note: this is an on-chart adaptive weighting mechanism built entirely with native MQL5 logic. It does not connect to any external server or third-party AI service.

Key Features

• Three-signal composite voting system (Trend / Momentum / Breakout)
• Self-adjusting signal weights based on real trading outcomes
• ATR-based Stop Loss, Take Profit, and optional trailing stop
• Spread filter and trading-session filter
• One position at a time - no grid, no martingale, no averaging
• Persistent weight memory across restarts (per symbol/magic number)
• Works on any symbol and timeframe supported by your broker

Input Parameters

• Magic Number - unique identifier for this EA's trades
• Time Frame - timeframe used for signal calculation
• Max Spread (points) - maximum allowed spread to open a trade
• EMA Fast / EMA Slow Period - trend signal moving averages
• RSI Period, RSI Bull/Bear Level - momentum signal settings
• Bollinger Bands Period / Deviation - breakout signal settings
• Learning Rate - speed of weight adaptation
• Min/Max Weight - clamping range for adaptive weights
• Vote Threshold - composite score required to open a trade
• ATR Period, Stop Loss (ATR), Take Profit (ATR) - risk management
• Lot Size - fixed trade volume
• Use Trailing, Trail Start (ATR), Trail Step (ATR) - trailing stop settings

How to Read a Trade

• Buy - the weighted sum of the three votes (Trend x weight + Momentum x weight + Breakout x weight) is above the Vote Threshold. Example: EMA fast above EMA slow and RSI above the bull level can be enough even if the Bollinger signal is neutral.
• Sell - the same composite score is below minus the Vote Threshold.
• Why it is not trading - the signals disagree (score stays between the thresholds), the spread is above Max Spread, the time is outside the session filter, or a position is already open (one trade at a time).
• Watching it learn - after each closed trade, the weights of the signals that agreed with a winner go up and those that agreed with a loser go down. A low Learning Rate means slow, stable adaptation; a high one reacts faster but is noisier.
• Tuning tips - raise the Vote Threshold for fewer, more selective trades; keep Min/Max Weight close together if you want all three signals to always matter. Compare settings in the Strategy Tester (12+ months, real ticks) and then on demo before going live.

Recommendations

Test the EA on a demo account first to confirm it fits your broker's trading conditions and your risk tolerance. Use the Strategy Tester with real tick data to observe how the adaptive weights evolve over time before trading live. As with any automated trading system, past behavior does not guarantee future results - always use appropriate risk management and position sizing for your account.

Risk & Requirements at a Glance

• One position at a time - no martingale, no grid, no averaging.
• Every position uses an ATR-based Stop Loss and Take Profit; optional ATR trailing stop with minimum-distance safety checks.
• Fixed lot size (Lot Size input), Max Spread filter and trading-session filter.
• A trade opens only when the weighted signal vote passes the Vote Threshold.
• Works on any symbol/timeframe your broker supports - test 12+ months with real ticks, then on demo.
• Defaults are a starting point, not tuned for your broker. Results are not guaranteed.

Frequently Asked Questions

Is this "AI" connected to ChatGPT or an outside server?
No. The adaptive part is a multiplicative-weights rule written in native MQL5. Everything runs inside your terminal; nothing is sent to any external service.

Does it use grid, martingale or averaging?
No. It holds one position at a time with a fixed lot size, and every trade has an ATR-based Stop Loss and Take Profit.

Why is it not opening any trades?
Usually the three signals disagree and the score stays inside the Vote Threshold, the spread is above Max Spread, the time is outside the session filter, or a position is already open. A lower Vote Threshold gives more trades but less selective ones - compare it in the Strategy Tester first.

How do I reset what it has learned?
The weights are saved per symbol and Magic Number. Using a new Magic Number starts the learning from the default weights.

Can I run it on several charts?
Yes - give each chart its own Magic Number. Each symbol/magic pair keeps its own trades and its own learned weights.

Is there a risk-percent lot, prop-firm limits or an MT4 version?
Not in this free version (it uses a fixed lot). These are available as a custom version - see below.

Need a Custom Version?

I can adapt this EA to your own rules - extra or different signals, risk-percent lot sizing, prop-firm daily loss / drawdown limits, a different session filter, or an MT4 version. Send me a private message here on MQL5 (or offer me a personal order via MQL5 Freelance) with your requirements, and I will reply within 24 hours with whether it is doable and the estimated scope. If you find this free version useful, please leave a review - it helps a lot.

More free tools from the same developer

• AI Confluence Signal Dashboard - an indicator that shows trend, momentum and trend-strength agreement on the chart, useful to see visually when signals line up before letting this EA trade.
• Ichimoku Trend Rider - a single-system trend-following EA, if you prefer one fixed, easy-to-read rule set instead of adaptive weights.
• AI CCI Reversal Trader - a mean-reversion EA for ranging markets, a different style to compare against this EA in the Strategy Tester.