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 proposes a nonlinear regression equation for forecasting prices from two prior observations. Its seven terms combine linear and squared price inputs, a difference representing momentum, a sine component intended to capture cycles, and a constant.…
The document presents a way to make trading forecasts more interpretable by translating fitted machine-learning models into algebraic expressions. Its example uses technical features derived from price bars, including returns, volatility, RSI, MACD, and…
This article describes incorporating an adapter-tuned GPT-2 model into a MetaTrader Expert Advisor for trading decisions. It compares three deployment approaches: converting the model to ONNX for in-platform inference, launching Python scripts from the EA,…
The document explains Artificial Cooperative Search, a population-based optimization method inspired by cooperative and predator–prey interactions. It maintains two candidate populations, selects one as predators and another as prey, shuffles prey solutions,…
This article develops code and a visual test bench for generating and inspecting random values from uniform, Gaussian, power-law, and Lévy-style distributions. Its focus is how distribution shape, finite bounds, and a selected center affect the likelihood of…
The article describes an MQL5 Expert Advisor that combines Q-learning, cross-pair correlation adjustments, and market features drawn from three timeframes. It represents market observations as an indexed state and scores a set of actions that includes…
The article explains Intelligent Water Drops (IWD), a population-based optimization method inspired by how water flow changes a riverbed. Candidate solutions act as drops that move through a search space. In the original method, drops favor paths with less…
The article applies category theory concepts to an MQL5 money-management example. It treats monoids as rule-based selectors for choices such as timeframe, lookback period, applied price, and indicator. Their selected values, together with a normalized…
The article explains simulated annealing as a stochastic method for searching complex optimization spaces, using the physical analogy of slowly cooling heated metal. It describes starting from an initial candidate, generating neighboring candidates,…
The article presents Volatility Adjusted Momentum (VAM), which scales price momentum by a volatility estimate and the square root of the momentum lookback. Positive and negative readings are treated as directional signals, with magnitude representing…
This article explains how a Deep Q Network can evaluate three trading patterns built from TRIX, a smoothed momentum indicator, and Williams Percent Range (WPR), an overbought and oversold measure. It outlines the value based reinforcement learning setup:…
This article shows how to inspect an ONNX model file inside MetaTrader 5 by implementing a Protocol Buffers reader and an ONNX-specific parser in MQL5. The reader handles tags, variable-length integers, length-prefixed data, fixed-width values, and unknown…
This article explains Growing Neural Gas, an unsupervised clustering method whose network topology expands and changes as it learns the distribution of input vectors. Starting from two neurons, it repeatedly finds the closest and second-closest neurons,…
The article introduces Mamba, a selective state space model proposed as an alternative to Transformer-based sequence modeling. It explains the motivation: attention has costs that grow with context length, while ordinary recurrent or convolutional models can…
The article describes Saplings Sowing and Growing (SSG), a nature-inspired population optimization method in which candidate solutions are treated as trees and parameters as branches. It presents an implementation that starts with randomly distributed…
The article explains Time-MoE, a decoder-only Transformer foundation model for time-series forecasting, and sketches how its components could be implemented in MQL5. It describes converting each time step into a token, embedding features with SwiGLU, and…
This article continues a description of a chaotic optimization algorithm for searching solution spaces. It combines logistic, sinusoidal, and tent maps with an initial broad search, weighted-gradient refinement, and a final local search whose scope narrows…
The article proposes organizing economic calendar observations with database tables and category-theory mappings. Its motivation is that calendar events differ in units, formats, geography, and meaning: some publish numeric indicators, while others are…
The article describes a two-stage workflow for measuring costs and using them to set more realistic triple-barrier labeling thresholds. An MQL5 script gathers broker-specific spread history, hourly spread averages, swap settings, and symbol details into a…
This article explains an attempt to adapt N-BEATS, a residual time-series forecasting architecture, for a forex Expert Advisor in MetaTrader 5. The model decomposes a series through stacked blocks that successively explain historical input and contribute…
This article describes a practical MQL5 adaptation of Multitask-Stockformer, a neural architecture for financial time series. It first outlines the framework’s use of discrete wavelets to split signals into low-frequency trends and high-frequency…
The article proposes using cross-validation predictions to identify training examples that a classifier repeatedly mislabels. A primary CatBoost model is trained on examples judged more predictable, while a second classifier learns across the wider dataset…
The article describes an implementation component for a segmental structured convolutional neural network intended to process noisy, nonstationary financial time series. Its design first decomposes a series into trend, seasonal, short-term, and residual…
The article demonstrates tuning an MLP regressor for Brent oil price forecasting with the derivative-free Nelder–Mead algorithm. It defines a target using a future close, evaluates candidate settings by time-series cross-validation RMSE, and searches…