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 presents Bag-of-SFA-Symbols (BOSS) as a way to classify market regimes by converting price windows into symbolic words. Each window is z-normalized to reduce sensitivity to price level, transformed with a low-pass Fourier representation to retain…
The article explains SSCNN, a neural architecture for multivariate forecasting that decomposes a series into long-term, seasonal, and short-term components plus residual information. A temporal attention normalization layer selects relevant observations for…
The article introduces OpenCL as a way to run computational workloads in parallel from MQL5, using a calculation of pi to illustrate the difference between a single CPU loop and an OpenCL implementation. It outlines the roles of vendor runtimes and SDKs, and…
This article describes a workflow for training a Twin Delayed Deep Deterministic Policy Gradient (TD3) agent in Python and deploying its actor network in an MQL5 Expert Advisor through ONNX. It presents the trading task as sequential decision-making:…
This article explains Charged System Search (CSS), a population-based optimization method inspired by electric charges and Newtonian motion. Candidate solutions are modeled as charged spheres. Their fitness determines their charge, while forces between…
The article explains Local Feature Selection (LFS), a classification method that chooses a potentially different subset of predictors for each training sample or local region. Instead of ranking features by their overall predictive value, it seeks features…
The article explores whether forecasting changes in RSI can provide a useful proxy for forecasting direction in Deriv’s Boom 1000 synthetic market. Using 100,000 one-minute observations, it compares a direct price-direction target with a target based on…
This article proposes a trading system that pairs a discrete Fourier transform (DFT) with a leaky integrate-and-fire spiking neural network. The DFT examines a rolling series of prices or indicators such as MACD and RSI, selects the strongest frequency…
This article introduces Prophet’s trend, seasonality, changepoint, and holiday components, then applies the model to hourly EURUSD data from MetaTrader 5. It defines the target as each bar’s high-to-low range, converts timestamps and volatility into the…
The article argues that predictive machine-learning models depend on meaningful inputs and labels, and that associations in historical data do not by themselves demonstrate cause and effect. It introduces causal inference concepts alongside the history and…
The article introduces genetic algorithms as population-based optimization methods and explains how candidate solutions are represented as chromosomes. It covers binary encoding of integer and floating-point attributes, including Gray coding to make…
This article describes an automated pipeline that combines Smart Money Concepts with a machine-learning filter. A Python process retrieves historical XAUUSD hourly bars, detects Order Block, Fair Value Gap, and Break of Structure events, labels events by…
The article explains partial information decomposition (PID), which separates the information two indicators provide about a target into redundant, unique, and synergistic parts. It describes three alternative redundancy definitions—I_min, I_MMI, and…
The article presents a binary classifier for predicting the direction of the next Renko bar on EURUSD. It converts five-minute price data into Renko bars using an ATR-based block size, then builds features from recent bar directions, directional streaks, and…
The article explains TimeFound, a Transformer-based foundation model designed to forecast time series across domains, including financial data. Its central preprocessing idea is multi-scale patching: segments at several resolutions are projected into a…
The article describes a proposed integration of ST-Expert, a graph-based mixture-of-experts approach, with Extralonger, a modular Transformer framework for time-series forecasting. It frames market data as both sequences and networks: assets or time points…
The article contrasts directional classification with regression for trading forecasts. A classifier predicts direction but not the size of a move, so hit rate alone can conceal the effect of gains and losses. The proposed regression targets measure future…
The article describes the Bat Algorithm, a population-based heuristic that models search agents as bats adjusting their positions through frequency, velocity, pulse rate, and loudness. It outlines initialization, movement, local search around the current…
This article surveys and implements market microstructure features for machine learning. Its OHLCV path estimates spreads and price impact using measures such as Roll, Corwin–Schultz, Amihud, and Hasbrouck. A separate path uses raw ticks to calculate per-bar…
The article explains how to combine numerical predictions from multiple trained models, with MQL5 implementations of several ensemble methods. It begins with a simple average, whose squared error is bounded by the average squared error of the component…
The article replaces a moving-average component in an earlier feedback-control trading design with a supervised statistical model of EUR/USD. It outlines a workflow that gathers daily OHLC data, labels future prices at a defined horizon, compares candidate…
The article presents an Optuna workflow for tuning financial machine learning models while preserving a financial data contract. It describes Bayesian parameter search with a TPE sampler, fold-by-fold evaluation using PurgedKFold, separate fitting and…
The article proposes SEAL, an experimental process for adapting a trading language model from trade outcomes. It stores predictions and results as training examples, assigns each example a weight based on factors such as directional correctness, model…
The article describes a reinforcement learning setup for trading that adds a Director to the usual Actor–Critic pair. The Actor selects actions, the Critic estimates their reward quality, and the Director classifies whether actions fit the learned strategy.…