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.
3,481 documents
This platform support exchange distinguishes continuing training from a saved StockRanker model from saving a model for a full training run. The user wants incremental training, such as adding trees or increasing the number of leaf nodes, but reports an…
This document presents two factor-based approaches to managed futures (CTA) investing. The first aims to time exposure to CTA strategies across changing market and macroeconomic conditions, using indicators from both technical and fundamental perspectives.…
The article explains a snowball structured product through observation dates, knock-in and early redemption conditions, and a capped return. It presents the payoff as conditional: investors may receive a preset return when the underlying stays within…
This research summary compares industry-level signals built from risk-adjusted returns and moving averages. It reports tests of Jensen’s alpha, Sharpe, Treynor, and Calmar measures across different lookback periods. The six-month Jensen measure behaved as a…
This overview introduces several major machine-learning approaches and explains their basic purposes. It contrasts linear regression for numeric prediction with logistic regression for classification, then describes neural networks as layered processing…
The article describes two ways large holders might try to recover from a losing stock position when selling could worsen market impact. With a relatively small holding, it claims an institution may sell into a gradual decline to accelerate a sharp drop, then…
This note describes several equity factors built from the distribution of individual trade values within minute bars. It uses distribution statistics such as quantiles, dispersion, skewness, and kurtosis, and examines correlations between trade value and…
The article examines whether valuation ratios predict future equity returns or offer a basis for market timing. Using long-run monthly U.S. stock market data, it compares the cyclically adjusted P/E ratio (CAPE) and conventional P/E with realized equity…
This research summary describes stock-selection factors built from Chinese public mutual-fund holdings. Because quarterly reports disclose only the largest positions and holdings data arrive infrequently, the study fills out recent quarterly holdings using…
This commentary examines retail traders who use post-market trading-seat disclosures, commonly associated with the A-share Dragon-Tiger list, as a signal to copy prominent investors. It argues that a disclosed seat is an outcome of a trader's decisions, not…
This forum post raises a replication problem involving a high-frequency stock return-variance factor described in a Haitong research report. The author says they implemented the report's first two calculation methods in SQL, but obtained results on a very…
This page outlines ways to retrieve foreign-exchange data from a market-data provider. It describes REST requests for historical candlestick data and current quotes, along with a WebSocket connection for subscribing to live quote and depth updates. The…
The document presents factor-category weighting as a simple neural-network problem. Gradient-based optimization can learn both the weights within categories and the weights across categories from cross-sectional observations. Under certain conditions,…
This report analyzes Chinese active equity funds using Fama-French and Treynor-Mazuy models to separate market timing, stock selection, and style allocation. It reports that positive stock-selection ability was more common than persistent market-timing…
The article introduces factor investing as portfolio construction based on characteristics associated with differences in risk and return. It outlines the development from the CAPM to the Fama–French three-factor and five-factor models, then distinguishes…
This report summary reviews stock-selection factors derived from individual trade records, including large-buy value share, buy-order concentration, intraday aggressive-buy activity, and informed trading near the close. It reports that orthogonalized factors…
This overview explains how neural networks combine neuron types and connection patterns, then surveys architectures including feed-forward networks, Hopfield networks, Boltzmann machines, autoencoders, recurrent networks, convolutional networks, and others.…
This document summarizes a commodity futures strategy based on carry and the futures term structure. It describes a directional rule in which the nearby, or main, contract is bought when its price is below the next contract and sold when it is above it. The…
This research note examines whether order book data can add stock selection information beyond factors derived from executed trades. It proposes three measures based on changes in net buy orders: the average rate of change, its volatility, and its skewness.…
This overview describes a commodity futures CTA strategy that ranks contracts using a combined score. It considers technical signals—cross-sectional momentum, volatility, open-interest changes, and term structure—alongside macroeconomic measures including…
This overview surveys a platform’s machine learning tools across classification, regression, ranking, and clustering. It names representative methods such as support vector machines, neural networks, gradient boosting, random forests, nearest neighbors, and…
This excerpt summarizes a two-week review of fundamental quantitative sector allocation and ETF rotation recommendations in China. It compares recommended overweight, benchmark-weight, and underweight industries with the broad Wind All A index, reporting…
The document explains how to inspect trained parameters in a Keras model that combines an LSTM layer, an auxiliary input, and several dense layers. It distinguishes trainable variables from the arrays returned by the model’s weight retrieval method, then…
This tutorial explains how to retrieve one field’s value from the day another field reaches its maximum within a rolling window. Its example seeks the trading volume on the day of the highest price over the preceding 20 days. The approach uses `m_imax` to…