Articles on machine learning in trading

In this article, we begin our exploration of the SSCNN framework — a modern architectural solution for time series analysis that combines accuracy, a structured design, and high computational efficiency. We will systematically examine its theoretical aspects, highlight the key differences from its predecessors, and begin the practical implementation of its basic components in the MQL5 environment.

We examine how a Hidden Markov Model (HMM) estimates latent market regimes while basing on observable price and indicator sequences. This is done by estimating the probability of state transitions. A Gated Recurrent Unit (GRU) network models time dependencies and keeps important information over several observations. In an Expert Advisor, HMM-based regime probabilities, can be merged with GRU-based sequence learning to better classify increments in accumulation, distribution, and momentum prior to their showing up in regular price confirmations.

We present a rule-set-aware calibration chain that turns the remaining risk budget into a calibrated sigmoid scale for position sizing. It computes a ceiling from stop loss pct and safety factor, back-solves w at a reference divergence, and flattens size progressively as the budget shrinks. The paper also clarifies where leverage caps must be applied in production: at the lots conversion, since risk-based sizing alone does not enforce max leverage.

The article shows how to tune ICA hyperparameters with a supervised evaluation pipeline and apply spectral clustering to time-lagged indicators. Cross-validation identifies the optimal number of clusters, which are translated into expected return and risk measures. These signals drive dynamic position sizing and stop-loss control, with surrogate models converted to ONNX and integrated into an MQL5 Expert Advisor.

The article proposes a synthesis of new technologies to overcome the limitations of classical indicators in market data analytics. It shows how language models and quantum encoding can reveal hidden market patterns that traditional methods overlook. The experiment confirms the value of new technologies and proposes an updated analysis methodology aligned with the current state of computational innovation.

We invite you to explore an innovative approach to forecasting time series with missing data using the GinAR framework. The article demonstrates the implementation of key components using OpenCL, which ensures high performance. In our next publication, we will take a detailed look at how to integrate these solutions into MQL5. This will help understand how to apply the method in practice in trading.

Net profit and win rate do not tell you if a strategy's edge is statistically real. This MQL5 toolkit analyzes return series built from price data or deal history and reports t‑statistics, p‑values, and confidence intervals using one-sample and Welch t‑tests, the Mann–Whitney U test, and volatility‑regime analysis to support evidence‑based trading decisions.

We build a rolling price channel by fitting the 0.1, 0.5 and 0.9 conditional quantile lines via IRLS with pinball loss, packaged as a reusable class and two MetaTrader 5 indicators. We verify in-sample coverage, examine quantile crossing, and compare the channel width with ATR, Bollinger and regression widths on matched horizons. Tests in the Strategy Tester show the edges are descriptive, while the normalized width works as a volatility/regime feature.

Bollinger Band mean reversion degrades in trending regimes when ADX is high and bandwidth expands. We separate direction from trade selection with a two‑stage meta‑labeling pipeline: a gradient‑boosted secondary classifier trained with PurgedKFold on band‑specific features (BBP, BBB, bandwidth regime) outputs action probabilities that drive probability‑based bet sizing. The MQL5 implementation loads the ONNX model and applies position sizing within a two‑EA architecture to filter low‑quality band touches.

The article describes the development of an MVP prototype for an autonomous trading bot for MetaTrader 5 that uses large language models (LLMs) via the OpenRouter API to analyze the market and make trading decisions. A Python script retrieves historical OHLCV data, sends it to an LLM for technical analysis based on support/resistance levels and Price Action patterns, and then automatically places orders with specified stop loss and take profit levels.

The article describes the process of fine-tuning a language model for trading based on real historical data from MetaTrader 5. The base model, which has only theoretical knowledge of technical analysis, is trained on a thousand examples of the real behavior of currency pairs (EURUSD, GBPUSD, USDCHF, USDCAD) over 180 days. After being trained using Ollama, the model begins to understand the specific characteristics of each instrument.