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 defines AI-based stock selection as using machine learning and deep learning to analyze financial and market information, identify patterns, forecast possible price moves, and generate investment suggestions. It describes inputs such as…
This configuration defines a Qlib workflow that trains a PyTorch feedforward neural network on Alpha360 features for CSI 500 stocks. It uses a forward close-price return label, robust feature normalization, missing-value filling, and cross-sectional rank…
This introductory survey organizes multi-agent reinforcement learning (MARL) by the relationships among agents’ objectives: fully cooperative, fully competitive, and mixed cooperation and competition. It explains that cooperative agents can be modeled…
This market review describes a year of changing styles in Chinese equities: cyclical and growth themes led during the recovery, followed by a shift toward value later in the year. It argues that a high earnings base and flattening growth expectations could…
This page introduces the exploration–exploitation problem in reinforcement learning: a learning agent must decide when to test actions that may reveal useful information and when to choose actions based on what it has learned so far. The lecture is…
The discussion explains how to repeatedly reduce a model’s factor list using feature importance. The example starts with 15 factors, removes the three with the lowest importance after a training and backtest run, and repeats the process until five factors…
The report compares linear, polynomial, Gaussian, and sigmoid support vector machine classifiers, along with support vector regression, for multi-factor stock selection. Its workflow extracts features and labels, preprocesses inputs, trains and tunes models…
The discussion compares quantitative research practices in China and overseas. It says Chinese investors more often try to predict stock prices directly, while overseas researchers may estimate missing observations, factor values, company revenue, or…
This introductory guide lays out five areas for developing quantitative trading capability: mathematics and statistics, programming, financial knowledge, strategy research, and practical testing. It highlights time-series and cross-sectional econometrics,…
Gold Dust proposes a robustness check for optimized trading systems that addresses the instability of financial-market statistics. Instead of optimizing one parameter set on one historical interval and forward-testing it, the method optimizes separate…
This BigQuant forum answer explains how to use a different feature set in each iteration of a rolling model workflow. The example loop updates training and test start and end dates for each rolling window, disables the backtest module during model runs, and…
This overview surveys recurring problems in quantitative trading, including unreliable or incomplete data, model risk, slippage and fees, parameter selection, real-time monitoring, and processing speed. It also discusses choosing machine-learning methods and…
The report describes an asset allocation approach that measures cycle states and uses machine learning to estimate the probability that assets will outperform one another. It reviews macroeconomic timing frameworks, then describes cycle factors derived from…
This tutorial surveys support vector machines, k-nearest neighbors, naive Bayes, and the perceptron as lightweight classifiers for relatively small datasets and feature sets. It explains SVM’s maximum-margin boundary, the role of support vectors, soft…
This configuration specifies a Qlib workflow for training the HIST model on Alpha360 features for CSI 300 stocks and evaluating its signals in a portfolio backtest. The data handler normalizes features, fills missing feature values, drops rows without…
This short discussion explains a label distribution chart used when building an AI trading strategy. The horizontal axis represents label identifiers, and the vertical axis shows how many observations belong to each label. The chart can reveal labels with…
This report presents a short backtest of a large-cap stock strategy attributed to a multi-agent AI trading bot and compares it with SPY. The stated test ran from January 4 to January 15, 2026, using Yahoo data and a universe of large technology and other…
This short introduction proposes hidden Markov models (HMMs) as a machine-learning technique for market analysis and timing. It says the article will explain the model, discuss similarities between HMMs and stock markets, and develop a multi-factor…
This configuration describes a Qlib workflow for training a KRNN model on China’s CSI 300 universe with Alpha360 features. It sets a historical data range, uses robust feature normalization and cross-sectional label ranking, and defines a two-day forward…
This training overview outlines a proposed workflow for using multimodal large language models in quantitative investing. It combines time-series databases, knowledge graphs, and reinforcement learning in a pipeline that connects data, signals, and portfolio…
This opinion piece argues that quantitative trading’s advantages extend beyond execution speed. It emphasizes systematic research and iteration: models can combine many fundamental and price-based signals, test relationships across historical data, and…
This overview compares machine learning and deep learning methods for predicting stock prices and trends. It describes LSTM and GRU recurrent networks for sequential data, CNNs for extracting patterns, bidirectional LSTMs, and deep neural networks. It also…
This document describes a MetaTrader 5 SuperTrend indicator that adapts its volatility bands and selected trade parameters using prior signal outcomes. It offers reversal and failed-breakout signal modes, with optional confirmation from RSI, tick-volume…
This report studies whether one industry’s past returns can help predict another industry’s future returns. It argues that information may spread gradually across related industries because investors cannot immediately assess every impact of a new shock. The…