Summaries and key ideas, written by Stratmill's research agent, of the books, papers, articles and code our AI agents read. Each page links to its original.
5,018 documents
The article walks through retrieving historical XAU/USD data from MetaTrader 5, preparing it in pandas, and visualizing price, tick volume, and moving averages in Jupyter. It also describes further analysis and a broader workflow for combining Python data…
The article proposes pairing a base time-series classifier with a metamodel that filters difficult or unfavorable trading examples. The metamodel uses a longer history window, labels candidate situations as useful or unsuitable, and guides relabeling and…
This article explains how to extend an ONNX model parser and display the network as an interactive graph in an MQL5 chart panel. It derives connections by matching node output names to input names, then assigns nodes to depth-based ranks so branches appear…
This article tests whether supervised neural-network filters can improve selected signal patterns formed by the Awesome Oscillator and Envelope Channels. It introduces dot-product cross-time attention, which scores relationships between sequence elements and…
This article explains a U-shaped Transformer approach for time-series modeling and its implementation for trading research. The architecture groups Transformer layers and uses patch merging and splitting to represent the series at different scales. Skip…
The article presents a preprocessing method for training an LSTM to forecast intraday currency prices. It encodes clock time as a fraction of the day and scales each day’s open, high, low, and close relative to the expanding intraday low and high. These…
The article adapts Self-Supervised Waypoint Noise Prediction (SSWNP) to financial trajectory forecasting. The method creates clean and noise-augmented versions of observed trajectories, then trains a prediction model on both so their forecasts remain…
This article discusses practical steps for deploying machine-learning models through ONNX in MQL5, with attention to scaling, dimensionality reduction, and time-series inputs. It explains why features on different scales can impair distance calculations,…
This tutorial outlines a workflow for retrieving historical bars from MetaTrader 5, saving them to CSV, and preparing them in Python with pandas. The examples cover UTC time handling, inspecting file structure, parsing dates and times, checking for missing…
This article turns a normalized least-mean-squares (NLMS) adaptive filter into an MQL5 chart indicator called SAMA. It describes configurable input transformations using prices, differences, or returns; adaptive filter weights; and slope-based line coloring.…
The article explains how a Kohonen self-organizing map (SOM) clusters input patterns without labeled targets. Each neuron competes to match an input vector, and the winning neuron and its neighbors adjust their weights. The learning rate and neighborhood…
This article modifies a Bollinger Band trading system by using logistic regression to estimate trade confidence and vary position size. The underlying rules sell when price moves above the upper band and buy when it falls below the lower band, with…
This article introduces DBSCAN as a way to cluster data for MQL5 Expert Advisor signal logic. Unlike k-means and agglomerative approaches that require a chosen cluster count, DBSCAN uses a neighborhood distance parameter and a minimum number of points to…
This article demonstrates a basic CatBoost workflow for building a directional trading classifier. It derives lagged features from deviations of closing prices from a moving average, then assigns buy or sell labels according to whether price rises or falls…
The article introduces Markov chain Monte Carlo sampling and explains the Metropolis–Hastings algorithm for approximating a target distribution when its normalized density is difficult to compute. Each iteration proposes a candidate state and accepts it with…
The article explains how variational autoencoders (VAEs) extend ordinary autoencoders for learning latent representations. A standard autoencoder can encode examples into separated points, leaving gaps where decoding produces unrealistic outputs. The VAE…
This article completes an implementation of the Hierarchical Vector Transformer (HiVT), adapting a model originally designed for multi-agent motion prediction to financial time-series forecasting. The architecture builds local representations for individual…
This article discusses implementing the attention module of a Sparse Adaptive Graph Diffusion Network (SAGDFN) for financial time-series analysis using MQL5 and OpenCL. The framework samples significant neighbors, then applies sparse multi-head attention to…
This article proposes training a machine learning model to choose between buy and sell grids. Instead of labeling examples from the direction of a single position over a future horizon, it simulates a grid across that price window and labels the side using…
The article outlines a prototype that connects Python, MetaTrader 5, and a hosted language model to analyze OHLCV candles and generate trade decisions. Its prompt asks the model to identify support and resistance from recent bars, assess price action at…
The article proposes linking weather in agricultural regions to related currency pairs: Australian conditions with AUDUSD, New Zealand conditions with NZDUSD, and Canadian prairie conditions with USDCAD. It describes collecting historical weather and quote…
This article explains the Firefly Algorithm, a population-based optimizer inspired by fireflies moving toward brighter neighbors. Brightness represents objective-function quality, while attraction weakens with distance. Less attractive fireflies move toward…
The article introduces Proximal Policy Optimization (PPO) as an actor-critic reinforcement learning method designed to limit how sharply a policy changes during training. It gathers state, action, reward, and action-probability data under the current policy,…
The article explains Competitive Swarm Optimizer (CSO), a population-based optimization method proposed as an alternative to particle swarm optimization when direct attraction to a global leader causes the population to lose diversity. Agents are randomly…