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.
4,194 documents
The document outlines an A-share stock screen combining a turnover range of 3% to 12%, at least one limit-up event in the prior 25 days, and year-over-year net profit growth above 20% and no more than 100%. Its rationale is to combine trading activity and a…
This Chinese Q&A explains why a genetic programming factor-mining workflow can report factors that pass a test yet return an empty factor collection. The answer says the run log shows that the global fitness test did not pass, so no factors were retained. A…
This post proposes a Chinese equity screen for stocks associated with the metaverse sector. Its final selection logic combines a volume-ratio band, a price ceiling, and a price-to-earnings limit, alongside sector and relative-ranking filters. The post also…
This short note outlines a way to represent the cost basis of stock holdings across market participants. It proposes estimating average holding cost by weighting transaction prices by traded quantities, then using daily costs and trading volume to estimate…
The document outlines a Chinese equity screening rule for metaverse-related stocks. It selects names with a positive daily return and a reported increase in position share above a threshold, interpreting the latter as a possible sign of institutional buying.…
This Chinese equity screen combines market, trading, and accounting filters. It focuses on Shenzhen main-board stocks with amplitude above 1, previous-day actual turnover between 3% and 28%, price-to-earnings ratios from 0 to 29.01, and price-to-book ratios…
This forum post asks how to test high-frequency equity factors built from minute bars. It cites examples from a China International Capital Corporation multi-factor series: the average squared correlation between the high and low prices of 50-minute bars,…
This brief Chinese-language assignment submission argues that factor quality matters more than the choice of predictive model: a strong factor is presented as the foundation of a strong strategy. It says the author ran a linear-regression strategy because…
This midyear financial engineering report reviews China’s equity market through the first half of 2020 and frames its outlook around two themes: timing the market by following trends and reviewing factor-based stock selection. Its summary describes a gradual…
This Chinese equity screen selects stocks with amplitude above one, ranks them by the day’s auction amount, and keeps the top five when their 2019 dividend ratio exceeds twenty-five percent. The rationale is to combine price movement and early trading…
This report introduces six ways to interpret machine-learning models and applies them to an XGBoost stock-selection model. It discusses feature importance, individual conditional expectation (ICE), partial dependence plots (PDP), a surrogate decision tree…
This report studies whether company patent data can help predict stock performance. It compares factors based on patent counts and types, including invention patents, utility models, and design patents. The report treats granted invention patents as the most…
This forum post flags a data-quality concern in the net profit year-over-year growth field for Beijing Stock Exchange listings. The author reports substantial differences between the values from two named market-data sources and shows many securities with a…
The document introduces support vector machines, outlining hard-margin classification for linearly separable data, soft margins for data with outliers, and kernel methods for nonlinear boundaries. It then applies an SVM with a radial basis function kernel to…
The document summarizes Chinese equity research on refining a traditional idiosyncratic volatility factor. The baseline measure is built from a Fama–French three-factor model. The report links volatility clustering in individual stock returns to persistence…
This Chinese equity research summary proposes refining a traditional idiosyncratic volatility factor by removing its cross-period cross-sectional dependence. The starting factor is derived from the Fama–French three-factor model. The report argues that…
This stock-selection proposal targets Chinese equities classified in the metaverse theme. Its initial conditions require prior-day turnover above 8% and a 2019 dividend payout ratio above 25%. The final version adds a price-to-earnings ratio below 20 and a…
This post outlines a Chinese equity screen that combines a daily amplitude threshold, a gain below 6% at the 9:25 observation, and a 2019 dividend ratio above 25%. It presents the idea as a blend of short-term activity and dividend-focused fundamental…
The document describes a long–short equity factor strategy that ranks stocks by a combined daily factor score. At the next market open, it holds the most positive and most negative scores, while positions are fully liquidated as soon as a stock falls outside…
The document outlines ways to construct an equity portfolio using dividend characteristics. One example selects stocks using forecast dividend yield, calculated from expected next dividends relative to price, then weights selected holdings by market…
The document compares the Fama–MacBeth two-stage procedure with generalized method of moments for estimating factor risk premia and testing asset pricing models. It outlines a typical workflow: estimate asset betas, form portfolios, estimate test-portfolio…
The document explains how to include a zero factor-exposure requirement in a mean-variance portfolio optimization. For asset weights collected in a vector and asset betas in another vector, portfolio beta is their dot product. Setting that product to zero…
The document discusses how to interpret regression coefficients from Fama-French three-factor and Carhart four-factor models applied to a portfolio of large DJIA companies. The portfolio has a statistically significant market loading, a negative and…
The document explains why a single equity risk model may not serve every portfolio task. Commercial models bring substantial research and data-cleaning expertise, but their design may prioritize objectives that differ from a user's needs. The choice depends…