All Blogs / Analytics & Forecasts

ONNX Strategy 1

7 April 2026, 19:00

Vitali Vasilenka

The ONNX (Open Neural Network Exchange) library has revolutionized automated trading. Previously, the path from idea to implementation was long and complex: a researcher would train a model in Python, and a development engineer would have to rewrite all the logic in a more productive language like C++.

ONNX Strategy 1: https://www.mql5.com/en/market/product/170202?source=Site+Profile#description

The EA Quantum Lab community presents ONNX Strategy 1, the first trading robot of its kind built on the full integration of the ONNX (Open Neural Network Exchange) format. This isn't just another Expert Advisor: it's powered by deep learning neural networks that improve on classic strategies without changing their logic. We've trained five unique models, and the first one is ready to change the way you think about gold trading.

Live Signal - https://www.mql5.com/ru/signals/2341254?source=Site+Profile+Seller

Start: $99 (0 copies) - Next: $149 (13 copies) - Next: $199 (20 copies) = Final Price: $290

📹 ➡️ https://youtu.be/u6Rl30nHHlI

Today, ONNX acts as a universal translator, allowing you to train a complex neural network in a familiar environment and then "feed" it to a high-speed trading terminal, such as MetaTrader 5, for instant decision-making.

Below, I'll explain how this process works and why it's becoming a standard in modern algorithmic trading.

What is ONNX and why is it important for traders?

ONNX (Open Neural Network Exchange) is an open format for representing machine learning models, developed by Microsoft, Facebook, and Amazon. It's not tied to a specific framework and allows trained models to be freely transferred between different tools and runtime environments.

For traders, this means a transition from "bulky" solutions to "lightweight" and high-performance strategies.

Timeframe: M15
Symbols: XAUUSD
Account Type: Any (ECN, Pro, Standard)
Leverage: from 1:100
Recommended Deposit: from $1,500

Our Telegram Channel ✅ https://t.me/EAQuantumLab

Key Advantages of ONNX in Trading

Speed (Microseconds): Calculations occur locally on your CPU or GPU. Latency is in the microsecond range, which is critical for scalping and high-frequency strategies.

Native MQL5 Support: The MQL5 language in MetaTrader 5 supports ONNX out of the box. A special OnnxRun() function executes the model directly within the terminal.

Independence: Your Expert Advisor will work even if the internet goes down, and you won't have to pay for API requests to external servers.

How does it work? System architecture

From a technical perspective, the process can be broken down into three key stages.

Stage 1: Model development and training in Python

The researcher prepares historical data (prices, volumes, technical indicators), builds, and trains the model (e.g., LSTM for price prediction).

Case study: One project used an LSTM model on 31,053 EURUSD H1 candlesticks. The data was split into 80% for training and 20% for validation. The model was trained using PyTorch on an NVIDIA RTX 4090 GPU.

Stage 2: Export to ONNX format

The trained model is converted into a standardized .onnx file.

Code example (PyTorch):

import torch

# Предположим, у нас есть обученная модель 'model' и пример входных данных 'dummy_input'
torch.onnx.export(model, dummy_input, "trading_model.onnx",
                  input_names=['input'], output_names=['output'],
                  opset_version=13)

Key parameter: opset_version must be version 13 or higher to ensure all modern model operators work correctly.

Stage 3: Integration into MetaTrader 5 (MQL5)

This is the most interesting stage. The generated .onnx file is added to the MQL5 Expert Advisor's resources.

Example logic in MQL5:

// Загрузка модели
long model_handle = OnnxCreate("trading_model.onnx");

// Подготовка входных данных (массив float)
float input_data[window_size][features];
// ... (заполнение данными из текущего рынка)

// Принятие торгового решения
if(output[0] > 0.5) OpenBuy();

Conclusion

The ONNX library is a bridge between the world of advanced data analysis in Python and the world of high-performance trading on the MetaTrader 5 platform. It allows traders-programmers to harness the full power of neural networks (LSTM, GRU, CNN) for price prediction without sacrificing execution speed.

By integrating ONNX into MQL5, traders can train financial market models in their preferred development environment (Python) and then trade with low network costs, high order book update speeds, and asynchronous order submission. This opens up new opportunities previously available only to select professionals and institutional traders.