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 adapts a machine-learning trading workflow for assets where long and short opportunities may be unevenly distributed. It labels only trades in a chosen direction, using a randomly selected holding period and a markup threshold to distinguish…
The article applies supervised learning as an additional filter for three Ichimoku and ADX-Wilder signal patterns that had previously performed poorly in forward-walk tests. Its model uses a spectral-mixture-inspired feature layer to represent oscillatory…
The article tests whether Federal Reserve economic data can improve forecasts of EURUSD. It compares ordinary daily market quotes with three FRED series: the broad dollar index, US bond interest rates, and expected inflation. After expressing the data as…
The article investigates whether price slope and its arctangent angle can improve machine learning forecasts. It notes that time-based slopes depend on how gaps such as weekends are represented, while its alternative ratio of price changes can become…
The article continues an introduction to representing domains and relationships with category theory concepts in MQL5. It discusses identity and composition, uses everyday examples to illustrate mappings, and presents template-based classes for domains and…
The article surveys loss functions available in MQL5 and explains how their properties affect neural-network training. It contrasts magnitude-based measures such as mean squared error and mean absolute error with classification losses such as categorical and…
The document introduces generative adversarial networks as a way to create synthetic financial observations when real samples are scarce, imbalanced, noisy, or sensitive. It explains the generator and discriminator roles and their alternating training…
The article proposes converting market features into binary sequences for technical analysis. Price direction, position relative to a moving average, movement strength, trading volume, and observations across timeframes are examples of inputs that can be…
This article presents a probabilistic framework for combining technical indicator signals. It treats a profitable entry as the outcome of interest and estimates each indicator’s historical hit rate, then uses a naive Bayes formulation to estimate the…
The article introduces monoids as sets equipped with a closed, associative binary operation and an identity element, then applies the idea to trading choices. Example decision sets include lookback length, chart timeframe, applied price, indicator, and…
This article explains quantization as mapping continuous or finely measured feature values into a smaller set of discrete levels. It describes how a histogram can help inspect a feature’s distribution and how cutoffs define bins, with values then encoded by…
The article explains Central Force Optimization (CFO), a population-based numerical optimization method inspired by gravitational motion. It initializes candidate probes in a bounded search space and assigns their objective values as masses. Better-scoring…
The article explains how to integrate an Optuna Bayesian hyperparameter search into an existing machine-learning training pipeline while retaining a scikit-learn search path. In the Optuna path, model parameters, sample-weight scheme, decay, and linearity…
The article describes a workflow for adding a machine-learning model to an existing MQL5 order block strategy that also uses Fibonacci analysis. It begins by translating the strategy into Python so its pattern detection, indicators, order handling, and risk…
This article explains particle swarm optimization (PSO) as a population search method. Each candidate solution has a position and velocity; its movement reflects its own best-known position and the best position found by the swarm, with random factors and an…
The article explains the Group Method of Data Handling (GMDH), focusing on its multilayer iterative algorithm (MIA), and describes an MQL5 implementation for building polynomial prediction models. MIA pairs input variables to create candidate partial models,…
The article builds on earlier neural-network trading experiments and describes testing small feedforward networks inside MetaTrader 5. It compares three input approaches, including candle-size information, price-shape measurements, and moving-average slope…
The article presents Stochastic Marginal Actor-Critic (SMAC), a reinforcement learning approach that adds latent variables to a stochastic actor policy. The latent representation is intended to make actions more expressive and exploration broader, including…
The article describes feature engineering for longer-range market forecasts by predicting the direction of a moving average rather than price directly, while also modeling when price and the average diverge. It compares logistic regression models using…
The article presents a multilayer perceptron as a transformation from market data to trading signals, separating the network's forward calculation from the algorithm that adjusts its weights. It discusses limitations of supervised, unsupervised, and…
This article builds and evaluates a supervised neural-network Expert Advisor that attempts to anticipate Bill Williams fractal turning points before the confirming candle appears. It describes choosing market inputs from historical OHLC movements, volume,…
The article diagnoses a gap between a language model’s ability to classify future price direction and its ability to make profitable trades. It describes balancing upward and downward examples, building a feature set from common technical indicators, and…
This article introduces principal component analysis (PCA) as a way to compress a dataset into fewer variables while retaining much of its variation. Its workflow standardizes variables, calculates their covariance matrix, obtains eigenvectors and…
The article translates four bet-sizing approaches from a Python module into an MQL5 system: probability-based sizing, sizing from forecast-price divergence, budget-constrained sizing, and reserve sizing learned from data. It outlines a layered design with…