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
Royal Flush Optimization (RFO) adapts genetic algorithms by representing each solution coordinate as a discrete sector rank rather than a binary string. It initializes a population of candidate solutions, selects stronger candidates, exchanges coordinate…
This article presents a practical implementation of DUET’s channel clustering stage for multivariate time-series forecasting. It transforms each channel with a Fourier method, compares frequency-amplitude vectors, normalizes the pairwise distances, and uses…
The article outlines a computer vision workflow for classifying EURUSD price direction from OHLC time series. It retrieves hourly data from MetaTrader 5, forms rolling windows, scales each window, and labels it according to whether the closing price is…
The article tests whether sequential bootstrapping reduces correlation among bagged trees trained on overlapping triple-barrier labels. It separates two changes often made together: reducing the number of sampled rows to reflect average uniqueness, and…
This article presents the practical architecture and training of StockFormer, a hybrid trading framework combining predictive coding with reinforcement learning. Its three predictive branches model relationships among assets and forecast market behavior at…
This article extends a brute-force method for finding trading patterns in currency data. It expands a polynomial signal’s inputs from close-open changes to several combinations of each bar’s open, high, low, and close, allowing coefficient combinations to…
This article describes a machine learning workflow for representing single-candle market data with UMAP, a nonlinear dimensionality reduction method. It builds ten features from price changes over a chosen horizon and from the candle’s open, high, low, and…
The document introduces deep double descent, a pattern in which test error can initially fall, rise near a model’s ability to fit the training data, and then fall again as training continues. It contrasts this with the usual overfitting picture and examines…
The document turns a rolling Fibonacci retracement level into supervised-learning targets. For classification, it compares a later price with the current price and a level calculated from a recent high-low range, assigning bullish, bearish, or no-hit labels.…
The document explains univariate time-series forecasting with ARIMA, whose autoregressive, differencing, and moving-average terms are set by the orders p, d, and q. It describes choosing these terms with PACF and autocorrelation plots and checking…
This article describes a data-driven way to choose RSI entry zones for a forex strategy. It groups RSI readings into ten bands, labels whether price rises or falls over a future horizon, and aggregates those outcomes by band to estimate likely direction. The…
This article updates Kohonen self-organizing map tools for algorithmic trading analysis. It replaces neuron identifiers tied to display pixels with grid indices, and adds feature normalization so inputs measured on very different scales can contribute to the…
The article adapts Conformer, a spatio-temporal continuous-attention transformer originally proposed for weather forecasting, to financial market modeling. It explains how the method uses attention across the same variables at different states, estimates…
This article explains how the discrete Fourier transform (DFT) represents a time series as frequency components and how a power spectrum summarizes the energy at each frequency. It outlines the fast Fourier transform as a more efficient way to compute the…
The article compares fixed-horizon labels with methods designed to represent trading decisions more directly. Triple-barrier labeling sets profit-taking and stop-loss thresholds around an event, plus a time limit; the label depends on which threshold is…
The article outlines an architecture for automating optimization of one trading strategy across several symbol and timeframe combinations. It organizes work as a hierarchy: a project contains stages, stages contain jobs, jobs contain optimization tasks, and…
The document explains Battle Royale Optimizer (BRO), a population-based optimization method inspired by battle royale games. Candidate solutions compete with their nearest neighbors: the fitter candidate wins, while the loser accumulates damage and moves…
In this interview, Automated Trading Championship winner Olexandr Topchylo discusses automated trading, investor accounts, and lessons from competition. He describes using neural-network-based Expert Advisors, but cautions that neural networks are complex…
This article describes the Monkey Algorithm, a population-based metaheuristic that searches a fitness landscape through repeated local improvement and broader exploration. Each candidate begins at a random position, climbs through local moves, and makes…
The article presents the Bison Algorithm, a population-based method for optimizing a continuous single-objective function. It divides candidate solutions into a large swarm group and a smaller explorer group. Swarm members move toward a target based either…
The article proposes a Chaos Attractor Oscillator that treats extreme price deviations as potential mean-reversion signals. Its outline combines a neural network trained on historical prices with an estimated dynamic attractor; the oscillator expresses the…
The document explains Coral Reef Optimization (CRO), a population-based search method that represents candidate solutions as corals occupying cells in a reef. It describes initializing the reef, generating candidates through crossover and mutation, letting…
This article presents an implementation of Brain Storm Optimization (BSO), a population-based optimizer inspired by idea generation in groups. Candidate solutions are assigned to clusters, and the algorithm forms new candidates by selecting cluster centers…
This article describes an adapter that connects a bet-sizing system to a trade-pyramiding engine in MQL5. It combines probability-based sizing, a concurrent-signal budget, and dynamic forecast-based bet sizing with an engine that adds progressively smaller…