Put your Investments on Autopilot.

The backbone of our trading software is an advanced machine learning model, designed to analyze vast amounts of market data and predict price movements with high accuracy. This algorithm goes beyond applying fixed rules; it learns and evolves as it is exposed to new data, continuously optimizing its performance.

Technical Details:

Historical Dataset: The model is trained on an extensive dataset spanning over 20 years of historical market data, encompassing price, volume, volatility, and technical indicators. The system processes over 1 million price bars across various timeframes, from minutes to daily data, ensuring a comprehensive and accurate market view.

Types of Data Used:

• OHLC (Open, High, Low, Close) prices across timeframes ranging from M1 (one minute) to D1 (one day).
• Trading volumes and order flow data.
• Derived technical indicators such as RSI, MACD, moving averages (SMA, EMA), and Bollinger Bands.
• Relevant economic events and historical macroeconomic data, including interest rate reports and inflation data.

AI Model Architecture:

We use a hybrid architecture comprising:

• Recurrent Neural Networks (RNNs), specifically Long Short-Term Memory (LSTM), are ideal for analyzing time-series data and identifying repetitive patterns in price data. This allows us to make accurate predictions for both short-term and long-term trends.
• Support Vector Machine (SVM) algorithms, which classify market behaviour based on key parameters, distinguishing between trending and sideways market phases.

Predictive Capabilities: The model can predict price movements with an average accuracy of 78% over a 4-hour horizon and up to 65% on a daily horizon. Backtests on historical data indicate an outperformance of 12% annually compared to the market benchmark.

Training Algorithm: The model utilizes a supervised learning approach. During training, the system was “fed” thousands of real market scenarios (e.g., bullish, bearish, and sideways trends) and optimized its internal weights to maximize prediction accuracy. We integrated Gradient Boosting into the source code to continuously enhance decision-making processes.

Evaluation Metrics:

• Accuracy: The percentage of correct predictions out of total predictions.
• Recall: The model’s ability to correctly identify significant market movements.
• F1-Score: A combination of accuracy and recall to provide an overall measure of the model’s effectiveness. In our case, the model achieved an F1-Score of 0.72 during evaluation tests.

Real-Time Updates: The model is recalibrated every 24 hours with fresh market data to ensure its predictions accurately reflect current conditions. This process is fully automated, requiring no manual intervention.

Advantages of the AI Algorithm:

• Continuous Adaptation: The system employs incremental learning techniques, enabling it to quickly adapt to new market data, prevent overfitting, and maintain its efficacy over time.
• Robustness Across Markets: Tested across multiple asset classes (forex, commodities, indices), the model has shown consistent predictive accuracy, with stable performance on currency pairs such as EUR/USD, GBP/USD, and commodities like gold and oil.
• Automated Optimization: We leverage an automated optimization framework using Grid Search and Random Search techniques, which continuously explores new parameter sets to maximize expected returns. This significantly reduces the risk of long-term strategy degradation.

This complex and well-engineered architecture allows our software to deliver reliable trading signals, minimizing human intervention and providing traders with a substantial competitive edge in the market.