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
This Chinese-language post outlines an equity screen that combines daily price movement with company profitability and a short-term price trend. Its initial rules require a price range of at least one unit, return on equity above 15% for five consecutive…
The document describes research that turns intraday stock volume data into factors for stock selection. It presents seven conventional volume distribution factors and six specialized ones. The methods include measuring minute volume through bucket entropy…
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…
The document presents a Chinese equity screening rule that combines trading activity, short-term price location, and a multi-year profitability filter. It selects stocks with turnover between 3% and 12%, an opening price within 5% of the 10-day moving…
The document describes a stock screen combining three conditions: daily price amplitude above 1, the average price above its five-day moving average, and return on equity above 15% for five consecutive years. It presents this as a blend of a short-term…
The post proposes a Chinese equity screening idea that combines a capital-activity measure, Bollinger Band position, and a high dividend ratio. It describes selecting stocks whose closing prices lie above the middle band and below the upper band, alongside a…
The article compares three consumer activity measures for publicly traded companies: estimated in-store visits, traffic to company and related websites, and interest in brands measured through search and social activity. It forms growth proxies by comparing…
The document outlines six assumptions used to simplify asset pricing theory, especially the Capital Asset Pricing Model: competitive and efficient markets, no taxes or transaction costs, risk-averse investors, shared expectations, normally distributed…
The research note describes factor-mimicking portfolios (FMPs) as an alternative representation of alpha factors, with conversion between the two under a given stock covariance structure. It argues that linearly combining alpha signals corresponds to…
This stock-selection proposal screens mainland Chinese main-board equities for a turnover rate between 3% and 12%, a robotics concept classification, free-float market capitalization below the stated threshold, and positive institutional-investor activity.…
This research summary examines whether Chinese listed companies’ earnings previews can add information to conventional equity factor models. Because previews arrive before earnings announcements and periodic financial reports, the report treats them as an…
This strategy ranks a universe of country equity ETFs using Shiller’s cyclically adjusted price-to-earnings ratio (CAPE), which compares inflation-adjusted prices with a ten-year average of inflation-adjusted earnings. At each year-end, it selects the…
This post describes an equity screen that selects stocks with a daily high-low range above a threshold, that did not close at the upper price limit on the previous day, and that have a positive price-to-earnings ratio. The proposed rationale is to seek…
The document proposes a Chinese A-share stock screen using RSI below 65, a positive product of daily price change and a five-day signed volume measure, rising moving-average conditions, and trailing price-to-earnings below 30. It explains the signals as a…
This research summary examines two responses to a prolonged value-factor drawdown: changing how the factor is constructed and forecasting when its returns may be stronger. It compares conventional cross-sectional valuation with time-series measures of a…
The post describes a proposed A-share stock screen using a turnover-rate range of 3%–12%, market capitalization below 10 billion yuan, and a profitability filter. It then adds a Bollinger Band condition, along with example formula and Python references. The…
This assignment asks learners to build a trading strategy using the standard deviation of turnover rate as a stock selection factor. It also specifies two exit rules: liquidate positions when the broad market falls by 4% consecutively, and close an…
This Chinese-language research roundup summarizes three studies from overseas literature. The first presents a numerical simulation approach for approximating American option values, with potential application to pricing domestic convertible bonds. The…
This proposed screen targets Shenzhen main-board stocks with turnover between 3% and 12%, positive product of the day’s price change and large-order net flow, price-to-earnings ratios from 0 to 29.01, and price-to-book ratios from 0 to 3.11. The article…
This stock-selection example screens Chinese equities using a turnover range, year-over-year net profit growth, and the prior day’s trading value. It then ranks qualifying stocks by institutional ownership and selects the highest-ranked names. The rationale…
The document introduces microstructure segmentation: using high-frequency price and volume data to divide trading into intervals with different trading characteristics. It argues that activity levels can correspond to different price-trend behavior and…
This research summary introduces APM, a stock-selection factor built from differences between morning and afternoon price behavior in Chinese A-shares. Its premise is that informed traders are more active in the morning, making morning price moves a stronger…
This article outlines an A-share stock screen that combines daily trading activity and company fundamentals. It selects stocks with turnover between 3% and 12%, market capitalization below 10 billion yuan, and no losses, then ranks candidates using revenue…
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…