Articles on machine learning in trading

This article shows how to configure a black-box model to automatically uncover strong trading strategies using a data-driven approach. By using Mutual Information to prioritize the most learnable signals, we can build smarter and more adaptive models that outperform conventional methods. Readers will also learn to avoid common pitfalls like overreliance on surface-level metrics, and instead develop strategies rooted in meaningful statistical insight.

This article builds the foundation layer of a twelve-part MQL5 market microstructure toolkit. It implements guarded math helpers (SafeDivide, SafeLog, SafeSqrt, SafeExp, SafeTanh), robust data validation (ValidateSymbolV2, SafeCopyClose), trimmed statistical estimators (robust mean var), a linear regression slope, shared structs, and an FFT. You compile a single include file that hardens indicators and expert advisors against silent numerical failures and standardizes data flow for later parts.

In offline learning, we use a fixed dataset, which limits the coverage of environmental diversity. During the learning process, our Agent can generate actions beyond this dataset. If there is no feedback from the environment, how can we be sure that the assessments of such actions are correct? Maintaining the Agent's policy within the training dataset becomes an important aspect to ensure the reliability of training. This is what we will talk about in this article.

In this article we present the implementation of a feature selection algorithm described in an academic paper titled,"FREL: A stable feature selection algorithm", called Feature weighting as regularized energy based learning.

The article presents a new approach to solving optimization problems by combining ideas from bacterial foraging optimization (BFO) algorithms and techniques used in the genetic algorithm (GA) into a hybrid BFO-GA algorithm. It uses bacterial swarming to globally search for an optimal solution and genetic operators to refine local optima. Unlike the original BFO, bacteria can now mutate and inherit genes.

Our next focus in these series on ideas that can be rapidly prototyped with the MQL5 Wizard, is a Custom Trailing class that uses the Blooming Filter. Trailing Stop systems are an optional but very resourceful part to any trading system that we want to explore more in these series besides the traditional Entry Signals.

Applying Python session boundaries to MQL5 broker timestamps misclassifies session membership by two to three hours on any non-UTC broker, corrupting session flags across the full backtest history. We implement CTimeFeatures.mqh, containing CRingBuffer and CTimeFeatures, with three EA-facing methods: Initialize (UTC offset capture and frequency gate configuration), Update (log return push to session-conditional ring buffers), and Calculate (cyclical encoding, session flags, and session volatility). The output is a flat double array drop-compatible with Python's get_time_features for sub-hourly, hourly, and daily timeframes.

The article delves into the implementation of a modified Forward Selection Component Analysis algorithm, drawing inspiration from the research presented in “Forward Selection Component Analysis: Algorithms and Applications” by Luca Puggini and Sean McLoone.

XLV is SPDR healthcare ETF and in an age where it is common to be bombarded by a wide array of traditional news items plus social media feeds, it can be pressing to select a data set for use with a model. We try to tackle this problem for this ETF by sizing up some of its critical data sets in MQL5.

A detailed guide on how to create a heat map indicator for MetaTrader 5 that visualizes the price distribution over time. The article reveals the mathematical basis of time density analysis, where each price level is colored from red (minimum stay time) to blue (maximum stay time).

In this article, we introduce a modified version of stepwise feature selection, implemented in MQL5. This approach is based on the techniques outlined in Modern Data Mining Algorithms in C++ and CUDA C by Timothy Masters.

Biogeography-Based Optimization (BBO) is an elegant global optimization method inspired by natural processes of species migration between islands within archipelagos. The algorithm is based on a simple yet powerful idea: high-quality solutions actively share their characteristics, while low-quality ones actively adopt new features, creating a natural flow of information from the best solutions to the worst. A unique adaptive mutation operator provides an excellent balance between exploration and exploitation. BBO demonstrates high efficiency on a variety of tasks.

We invite you to explore the Actor-Director-Critic framework, which combines hierarchical learning and a multi-component architecture for creating adaptive trading strategies. In this article, we take a detailed look at how using the Director to classify the Actor's actions helps to effectively optimize trading decisions and improve the robustness of models in financial market conditions.

Batch normalization is the pre-processing of data before it is fed into a machine learning algorithm, like a neural network. This is always done while being mindful of the type of Activation to be used by the algorithm. We therefore explore the different approaches that one can take in reaping the benefits of this, with the help of a wizard assembled Expert Advisor.

In this article we present yet another custom MQL5 Signal Class that we are labelling ‘CSignalBTreeBayesian’. We are marrying the algorithm of a balanced tree with a neural network that is built on Bayesian principles to formulate yet another custom signal testable independently or with other signals thanks to the MQL5 Wizard.

We bridge Python-native artifacts to MQL5 for tick-accurate CPCV backtesting. The export script converts the ONNX model, calibrator, feature spec, and path masks to flat files, while the expert advisor rebuilds features, performs ONNX inference with calibration, and trades on real ticks. The Strategy Tester runs each combinatorial path, and Python aggregates per-path equities into a path Sharpe distribution to assess robustness after spread, slippage, and commission.

This article implements a Fenwick Tree (Binary Indexed Tree) for volume-aware money management inside an MQL5 Wizard Expert Advisor. We structure cumulative volume in O(log n) and apply four scaling modes—linear, conservative, aggressive, and mean-reversion—optionally gated by a lightweight 1D CNN. Practical tests compare the algorithm alone versus the CNN‑filtered approach to illustrate adaptive lot sizing and risk control under varying volume topologies.

This article starts the MMAR pipeline on EURUSD M5 data. We load market data via the MetaTrader5 Python API and run partition-function analysis with non-overlapping intervals to test for multifractal scaling. The result is an evidence-based decision on fractality, a prerequisite for building MMAR and for choosing whether to proceed beyond GARCH.

Downloading international monetary fund data in Python: Mining IMF data for use in macroeconomic currency strategies. How can macroeconomics help an ordinary and an algorithmic trader?

Building on the partition function analysis from Part 1, this article deepens the theoretical foundation before completing the analytical pipeline. We first give a full treatment of the Hurst exponent: what it measures, what it implies about market memory, and why it matters for the MMAR. This is followed by an intuitive exploration of multifractal spectra and what f(α) reveals about volatility heterogeneity. We then move to implementation: extracting the scaling function τ(q), estimating H via R/S analysis, and fitting the multifractal spectrum across four candidate distributions. By the end, we have the complete parameter set needed to construct the MMAR process in Part 3. Part 2 of an eight-part series.

RSI accumulates losses in trending conditions by firing at every threshold crossing regardless of market regime. A Random Forest secondary classifier trained on 12 contextual features — RSI momentum slope, EMA50 trend velocity, ATR-normalised trend stretch, and nine others — filters raw signals and scales position size by classifier confidence on EURUSD H1. Results compare plain RSI, meta-filtered RSI, and bet-sized RSI across a 16-month out-of-sample period with per-trade metrics and drawdown diagnostics.

This work presents an end-to-end pipeline: collect MetaTrader 5 data, engineer entropy/volatility/trend features, train a PyTorch classifier, and expose predictions through a Flask API. An MQL5 EA posts rolling prices each tick, receives probability and regime, and applies adaptive position sizing and stop distances. The result is a clear recipe for integrating ML inference with MetaTrader 5.

With the multifractal parameters from Part 2 in hand, this article builds the full MMAR process. We construct the multiplicative cascade for trading time, generate Fractional Brownian Motion via Davies-Harte FFT, and combine both into X(t) = B_H[theta(t)]. A 100-path Monte Carlo simulation produces the volatility forecast, which we then pit against GARCH on the same EURUSD M5 data. Does Mandelbrot's fractal architecture outforecast Engle's conditional variance framework? Part 3 of a eight-part series leading to a native MQL5 library and Expert Advisor.