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 project note describes iterative construction and evaluation of equity factors using amount, turnover, price, market capitalization, and price-to-book data. Candidate expressions combine standardized ratios and rolling statistics, including volatility,…
This article describes a reproduction of residual network experiments on CIFAR-10, focusing on two changes from an earlier run: image augmentation and shortcut design. It pads training images by four pixels on each edge and randomly crops them back to the…
This document outlines interface and configuration changes for upgrading Freqtrade strategies from version 2 to version 3, especially when adding short trades or leverage. It maps older buy and sell signals to entry and exit terminology, including renamed…
The document summarizes a study of stock price movement prediction in China’s A-share market. It describes combining five data types: historical trading data, technical indicators, company announcements, news, and Baidu search activity. The researchers…
This overview introduces several machine learning approaches used in stock strategies: deep neural networks, convolutional networks, LSTMs, gradient-boosted trees, and a ranking model built around GBDT. It briefly describes what each model can learn or…
This short note introduces data preprocessing as a step that can help machine-learning and deep-learning estimators work with raw feature vectors. It points readers to scikit-learn’s preprocessing module, which provides transformation classes and functions…
The article contrasts discretionary stock trading with quantitative approaches that analyze large datasets for statistical relationships. It describes using price and volume alongside less conventional inputs, such as online search activity and social media,…
This short forum exchange raises a practical concern about transparency in visual quantitative research tools. A user asks for fuller descriptions of how drag-and-drop modules calculate outputs, especially whether feature-generation steps first group data by…
This research note describes a daily equity alpha approach that uses supervised neural networks to generate many candidate factors from designed factor units. It then transforms the original inputs into weak, mutually orthogonal factors and weights them…
The document introduces ALGLIB, a numerical analysis library ported for use with MetaTrader. Its listed capabilities include linear algebra, optimization, interpolation, Fourier transforms, differential equations, statistical tests, and data analysis methods…
This study adapts SinGAN, a multiscale generative adversarial network trained on one sample, to generate financial return sequences. Its pyramid trains progressively across scales: coarse layers represent broad, low-frequency structure, while finer layers…
This edited seminar transcript surveys a Chinese quantitative research program on deep learning, then focuses on AlphaNet, a neural approach to stock factor discovery. It frames a multi-factor process as factor generation, factor combination, and portfolio…
The document outlines a way to run multiple parameterized tasks in BigQuant, a low-code platform where a visual workflow is represented as a graph of modules. A parallel mapping function can dispatch different inputs to a function, and the same pattern can…
The document describes deploying the small Kronos pretrained model on a quantitative research platform to forecast five-minute candlestick data for stocks. Kronos represents OHLCV sequences as hierarchical discrete tokens using a specialized tokenizer, then…
This research summary compares Naive Bayes, linear discriminant analysis (LDA), and quadratic discriminant analysis (QDA) as supervised methods for multi-factor stock selection. The described process extracts and preprocesses features and labels, retrains…
This paper-reading digest summarizes several computer vision methods: InstanceFCN for generating object instance masks from position-sensitive score maps; a two-stream residual network that combines full-resolution detail with pooled features for…
This event outline presents a learning series on using large language models and AI agents in quantitative research workflows, with examples centered on VeighNa strategy development. Topics include agent and tool interaction, model tool calls, MCP, task…
This note argues that quantitative strategy results depend on the combination of input factors and predictive models. It says that different factors can produce substantially different backtests when paired with the same model, and that a given factor can…
This document describes a native MQL5 library for creating, training, and running feedforward neural networks inside MetaTrader 5. It outlines dense matrix operations, activation functions, layers with dropout and gradient clipping, optimizers, learning rate…
This research summary describes a model for estimating whether a company will meet or exceed analyst expectations in its next reporting period. Its premise is that earnings delivery has persistence: stocks that beat expectations in one period are more likely…
The article introduces sentiment analysis as a way to gauge crowd opinion from online posts and exchange data, with cryptocurrency as its main application. It describes classifying social media messages as positive, neutral, or negative, and mentions social…
The document raises a measurement question about Rank IC, a rank-based measure of association between model predictions and subsequent returns. The author reports that Rank IC computed on a custom training set differs substantially from the value produced by…
This configuration describes an order-execution backtest using five-minute market data and an order file. The main strategy uses a recurrent network with a PPO policy, a categorical action interpreter, and a state interpreter that supplies recent intraday…
This research summary examines generalized linear models as ways to combine stock factor exposures into forecasts of future returns. Its workflow includes extracting features and labels, preprocessing features, forming training sets, and retraining on a…