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 describes an MQL5 forecasting model for the FTSE 100 that uses the index and ten large constituent stocks as inputs. It standardizes the historical input series with means and standard deviations, then fits multiple linear regression to predict…
The article explains Chimera, a two-dimensional state space model that represents dependencies across both time and variables in multivariate series. It describes temporal and feature-axis state updates, cross-dimensional information flow, discretization…
This article explains how attention mechanisms help neural networks weigh relationships among sequence elements. It traces the idea from encoder-decoder attention for machine translation, where hidden states are scored for relevance to the current decoding…
This article describes a custom trailing stop intended for noisy, fast-moving markets. Its first component applies a Hampel filter to a recent price window: the median and median absolute deviation define a range, and prices outside that range are treated as…
The article presents LSEAttention, a Transformer modification intended to reduce attention collapse and numerical instability in multivariate time-series forecasting. It uses the Log-Sum-Exp reformulation to stabilize Softmax calculations against overflow…
The article describes a trading model that combines a ResNeXt-based market encoder with multiple task-specific outputs. One output acts as an agent to generate trade parameters, while another estimates the probability of future price direction. Sharing the…
The article introduces the Mask-Attention-Free Transformer (MATF), originally developed for 3D point-cloud segmentation, and describes an interpretation for trading models. MATF addresses slow Transformer training attributed to poor initial object masks by…
This article explains an interpretation of the Artificial Bee Colony (ABC) algorithm, a population-based method inspired by how honey bees search for food sources. Candidate solutions represent locations in a multidimensional search space, and the objective…
This article presents a NumPy-inspired utility class for MQL5, intended to make mathematical and data-processing code easier to translate from Python. It surveys vector and matrix constructors, arithmetic operations, mathematical and statistical functions,…
This article describes the property-aware attention component of an Atom-Motif Contrastive Transformer for forecasting market patterns. It treats candlesticks as atomic observations and larger chart patterns as motifs, then uses contrastive learning to…
The article tests whether forecasting performance differs across 11 EURUSD time frames, using equal-sized samples of historical and future closing prices and a 20-step forecast horizon. It reports that models had lower errors on monthly and hourly data than…
This article builds a Multifractal Model of Asset Returns (MMAR) from previously estimated scaling parameters and compares its volatility forecasts with a GARCH(1,1) model. It creates multifractal trading time with a normalized multiplicative cascade, then…
The article describes an MQL5 tool that tests whether trade sequence, position size, and market context help predict losses. It encodes each closed deal with eight features and trains a small feed-forward neural network using stochastic gradient descent. A…
This article explains a stepwise feature-selection method based on mutual information, designed to identify predictors that relate to a target while avoiding redundant features. For continuous data, it contrasts density estimation with Parzen windows, whose…
This article presents an iterative pattern-search tool for trading, including a finite, modified Fourier-series polynomial as an alternative family of functions for describing price behavior. Its coefficients are generated through several random schemes, and…
This article describes a trading-oriented implementation of the PSformer time-series architecture. It reviews parameter-sharing blocks and spatial-temporal segmented attention, then explains how the encoder handles sequential market data. Because each bar…
This article walks through configuring Windows Subsystem for Linux 2 on Windows to run a local large language model. It covers installation choices, starting and managing Linux distributions, moving a distribution to another drive, and setting memory and…
This tutorial describes how an MQL5 program can communicate with Google’s Generative AI API through WebRequest. It covers API keys, request formatting, parsing JSON responses, extracting generated text, and saving that text to a file when the Expert window…
The article explains structural break tests as features for identifying changes in market behavior. It covers the Chu–Stinchcombe–White test for departures from a no-trend baseline, Chow-type Dickey–Fuller testing for a single unknown break, and SADF and…
The article examines whether apparent market randomness hides recurring hourly structure. It contrasts conventional autocorrelation of hourly price differences with seasonal autocorrelation, calculated after retaining observations from selected hours across…
The document introduces fuzzy sets as a way to represent degrees of membership between fully excluded and fully included. It defines membership functions, linguistic variables, terms, and defuzzification, using age categories to illustrate how a value can…
This article explains disagreement-based exploration in reinforcement learning. An ensemble of forward dynamics models predicts the next environment state from the current state and action. The variance among their predictions becomes an intrinsic reward:…
This article continues the implementation of the Artificial Bee Hive Algorithm, a population-based optimization method inspired by bee foraging. Agents occupy different behavioral states, including novice, experienced, search, and food-source roles. Their…
This article revisits MACD and On-Balance Volume signal patterns and describes using supervised learning to improve their interpretation. It outlines a one-dimensional convolutional network whose kernel sizes and channel counts are designed in relation to a…