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 document introduces Joint Recurrence Quantification Analysis (JRQA) as a way to examine whether two time series revisit their own past states at the same time indices. It distinguishes this from cross-recurrence analysis, which compares states across the…
The article develops a sequential detector for possible changes in return behavior and describes its implementation as a MetaTrader indicator. It converts prices to log returns, standardizes each observation using a strictly historical rolling window, and…
This article presents four bet-sizing methods for turning trading signals into positions: probability-based sizing, dynamic sizing from forecast-price divergence, budget-constrained sizing, and reserve sizing fitted to concurrent position imbalance. The…
The article explains Cross-Covariance Attention (XCA), which computes relationships across feature channels rather than across tokens. This changes attention's scaling with sequence length from quadratic to linear, making it suited to long sequences when the…
The article shows how Numba just-in-time compilation can accelerate computationally intensive Python tasks while keeping a Python-based workflow. It explains that code intended for Numba’s fast nopython mode should avoid unsupported objects such as Pandas…
The article introduces k-means clustering and contrasts it with hierarchical clustering and alternatives such as k-medians, k-medoids, Jenks natural breaks, and fuzzy clustering. It focuses on k-means++ initialization, which chooses initial centers in a more…
This article introduces K-nearest neighbors as a non-parametric method for classification and regression, then walks through a classifier implemented in MQL5. For a new observation, the method calculates its Euclidean distance from labeled examples, selects…
This article compares a EURGBP daily moving-average crossover system with an AI-assisted version. The baseline combines exponential moving averages, a stochastic oscillator, and ATR-based exits. The proposed method trains models on earlier data to estimate…
The article introduces Mamba, a selective state space model architecture, as an approach to processing long financial time series with linear sequence-length scaling. It explains the hidden-state formulation, input-dependent selection of information, and…
This article walks through a feed-forward neural network in MQL, from the weighted sums and biases of individual neurons to a network with two hidden layers. It uses hyperbolic tangent activations in the hidden layers and softmax outputs for three trading…
This article introduces dimensionality reduction as a way to manage large, correlated sets of trading indicators. It presents truncated singular value decomposition (SVD), which projects centered data onto a smaller number of components, and non-negative…
The document examines the tension between non-stationary price levels, which retain historical information, and integer differencing, which can discard that memory. It introduces long-range dependence and fractional differentiation as a way to seek…
The document introduces Recurrence Quantification Analysis (RQA) as a way to measure repeated patterns in time series. It explains time-delay embedding, distance calculations between reconstructed states, and thresholding those distances to form a recurrence…
The document explains the Dream Optimization Algorithm (DOA), a population-based method that could be used to tune trading-system parameters. It divides candidate solutions into groups with different numbers of dimensions to modify. During exploration, group…
This article describes a cross-market trend-following method that combines each market’s own momentum with momentum from markets that tend to lead it. It first scales price changes by estimated volatility, then forms moving-average crossover oscillators at…
The article presents ACEFormer, a neural architecture for forecasting financial time series, especially noisy and irregular high-frequency data. Its pipeline first smooths inputs with convolutional filters, then applies an adaptive ensemble empirical mode…
This article describes an MQL5 program that connects to an AI chat service and presents a multi-turn conversation in a scrollable chart-based interface. The interface design includes timestamps, conversation history, scrollbar display modes, mouse-wheel…
The article examines why Go-Explore’s two-stage process—exploring states and then training a policy from collected examples—becomes harder as a trading history grows. It reports that random action selection produced a profitable pass on a one-month training…
This article explains FEDformer, a Transformer architecture for long-horizon time-series forecasting that combines seasonal-trend decomposition with frequency-domain processing. Its Fourier version randomly selects a subset of frequency components, while its…
This article describes using the Mantis time-series architecture as a market representation module within a trading agent. It processes separate market-data channels and their first differences, extracts local features with convolutions, pools them into…
This article extends an online analytical processing framework to explore MetaTrader optimization results alongside trading reports and price data. It describes representing records in multidimensional cubes so users can filter, aggregate, sort, and compare…
This article explains how evolutionary methods, especially genetic algorithms, can optimize Expert Advisor parameters and predictor combinations. It outlines how candidate solutions are represented as chromosomes, scored with a fitness function, selected,…
The article describes an ML workflow in which Python trains a model while MQL5 prepares its inputs and runs inference through ONNX. It represents observations and features as matrices, then applies normalization and, optionally, principal component analysis…
The article applies a secondary machine-learning classifier to RSI crossover signals. RSI remains responsible for trade direction, while a Random Forest estimates whether each signal is likely to succeed using 27 contextual features measured at the signal…