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
This proposed system uses a large language model to generate and revise candidate price-and-volume factors for a cross-sectional long-short portfolio of USDT perpetual contracts. It screens liquid instruments, classifies market conditions using Bitcoin…
The article explains CorrGAN, a generative adversarial network designed to create synthetic financial correlation matrices. The motivation is that historical market data can be costly, restricted, biased toward the events that occurred, and sparse in extreme…
The release notes describe additions to a financial machine learning library, including time bars and information driven bars, structural break tests, market microstructure measures, entropy estimators, volatility estimators, clustering, dependence metrics,…
This tutorial presents preprocessing and labeling methods for supervised trading models. Fractional differentiation is used to make price features more stationary while retaining more of their historical dependence than ordinary differencing may preserve.…
This lecture series surveys advanced pairs and statistical arbitrage methods. Topics include distance-based pair selection and dependence measures, cointegration with mean first-passage time for choosing trading boundaries, PCA strategies, machine learning…
This March 2019 research update summarizes a project report on applying financial machine learning methods to trend-following and mean-reverting strategies. The report combines event-based sampling, the triple-barrier labeling method, and meta-labeling, and…
This paper describes the motivation and design of a Python research package intended to make methods from financial machine learning easier to implement and study. It frames Lopez de Prado’s work as a research process built around data preparation, sampling,…
This article presents an unsupervised learning framework for narrowing the search for equity pairs that may exhibit mean reversion. It first applies principal component analysis to asset returns to represent shared risk exposures, then uses density-based…
This document describes a pairs trading method that uses a two-state Markov regime-switching model to assess whether spread deviations may reflect a persistent change rather than temporary mean reversion. The proposed signal combines the estimated regime and…
This article presents the generic non-parametric representation (GNPR) distance for comparing time series using both distributional and dependence information. The motivation is that correlation or other familiar similarity measures can make series appear…
This document explains history-weighted, or partial sample, regression as a way to make predictions from observations judged relevant to a new input. It defines similarity using negative Mahalanobis distance and informativeness by how far an observation lies…
This article compares time, tick, volume, and dollar bars as ways to organize market data for machine learning. Time bars use fixed intervals; tick and volume bars use trade counts or traded quantity; dollar bars use traded value. The proposed rationale for…
The article explains why ordinary bagging can be problematic for financial labels. In event-based datasets, labels may share underlying returns, so observations are not independent. It introduces concurrency to describe overlapping information and uniqueness…
The document reflects on Quantopian’s closure through the strengths and constraints of its research platform. It describes benefits for systematic traders, including a team environment for exchanging ideas, training, research technology, peer feedback, and a…
This installment in a deep learning for trading series explains why GPU hardware can speed up the matrix operations common in neural network workloads. It outlines a Windows setup path for using Keras with TensorFlow from R: check hardware compatibility,…
This introductory article asks whether deep learning can be useful for market forecasting and outlines the practical work involved. A trading researcher must frame the prediction as a suitable task, scale inputs, choose a network structure, tune model and…
This article uses k-means clustering to group daily GBP/JPY candles according to their high, low, and close relative to the open. It examines whether particular candle clusters tend to follow one another and whether returns after each cluster differ. The…
The document explains how the Graphical Lasso estimates a sparse inverse covariance matrix from stock data. After scaling its off-diagonal entries, the method derives partial correlations, which describe the relationship between two stocks while accounting…
This tutorial describes how to connect the R statistical environment to Zorro, allowing a Zorro script to send market data to R, run R computations, and retrieve results. It outlines configuring the R installation, starting and checking an R session, and…
The article argues that AI makes it easy to generate and test trading rules, but that speed also encourages data mining. Repeatedly changing parameters, filters, timeframes, or asset universes amounts to many hypothesis tests; a strong historical result can…
The article argues that language models are unreliable for discovering trading edges because of three problems: conventional trading advice dominates their training data, models struggle to retrieve the latest value after repeated updates, and their…
This tutorial demonstrates a basic feed-forward neural network workflow for classifying the direction of hourly foreign exchange price changes. It constructs features from hourly changes in closing, high, and low prices, along with distances among those…
This guide introduces the perceptron, a basic neural network model for binary classification. It outlines activation functions and learning, then demonstrates how weights and a bias can be updated from classification errors. Examples use iris flower…
The article outlines a framework that groups daily candle patterns with k-means, then tests whether particular clusters support long or short trades. Its sample features are the day’s high, low, and close relative to its open. Historical observations are…