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
The document presents the Brinson framework for explaining portfolio returns through the benchmark, active asset allocation or timing, security selection, and their interaction. The benchmark reflects long-term asset-class weights and passive returns. Timing…
This forum post reports that two BigQuant notebooks produce different numbers of extracted features even when no filters are applied. The author observes that one notebook appears to include Beijing Stock Exchange data while another does not, and questions…
This glossary introduces common measures for evaluating investments and trading strategies. It explains cumulative, annualized, and monthly returns; benchmark-relative performance; win rate and reward-to-risk; information ratio; standard deviation; alpha and…
This Chinese-language research summary examines whether stocks removed from the CSI 300 later rebound. It compares cumulative excess returns for additions and deletions around index changes, measuring performance from 260 trading days before a change through…
This report presents a fundamental quantitative stock-selection framework for China’s lithium battery industry. It first reconstructs the supply chain into battery makers, material suppliers, and upstream cobalt and lithium resources, arguing that…
This report summary presents a framework for converting intraday or other medium-to-high-frequency signals into factors intended to predict returns over a monthly horizon. It separates factor construction into signal generation, aggregation from intraday…
The discussion explains how to change a StockRanker strategy’s initial buying schedule when starting a live simulation. The questioner says the strategy gradually adds positions, leaving the portfolio underinvested at launch, and wants the simulation’s first…
The document describes a strategy that ranks Chinese A-share market concepts by their recent equal-weighted stock returns, then buys stocks in winning concepts and shorts those in losing concepts. Portfolios are rebalanced monthly, with formation and holding…
This platform discussion addresses a reporting difference between older and newer versions of a StockRanker workflow. A user says the older version displayed feature importance and NDCG after a backtest, while those results were not visible in the newer…
This report summary describes a Chinese equity industry-rotation approach that scores sector conditions using financial data from regular reports, earnings forecasts, and earnings bulletins. It evaluates individual accounting measures, then combines…
This tutorial outlines a long and short strategy that uses short and long simple moving averages to identify trend changes. A bullish crossover triggers a long signal, while a bearish crossover triggers a short signal. It describes applying the approach to…
This BigQuant example builds a long-only Chinese stock portfolio from a filtered universe of main-board shares. It excludes suspended stocks and, in its feature filters, keeps non-ST stocks listed for more than a year, with positive trailing earnings and…
This platform discussion describes a data-ordering problem in a quantitative feature extraction workflow. The author reports that extracted rows and the original data may be separated or misaligned instead of being grouped and ordered by stock. Applying a…
This discussion explains how a quantitative platform interprets factor weights and describes its StockRanker as a learning-to-rank model based on a gradient-boosted decision tree structure. The response says the model does not require independent factors,…
This research compares ways to construct and transform analyst-consensus forecast factors for stock selection and industry rotation. It examines expected return on equity, net profit, net-profit growth, and a two-year compound-growth measure. The approaches…
This BigQuant forum exchange addresses two questions about accessing A-share financial statement data. The first asks why requesting a prior-year return-on-equity field by adding a suffix to the current field name causes an error. The reply explains that the…
This note outlines a Chinese stock screen requiring daily price amplitude above 1%, a stock code beginning with 60, and large-order net volume above 0.05 for at least three consecutive days. The proposed rationale is that a wide daily range indicates…
This strategy concept uses gradient-boosted regression to score stocks for leadership over a chosen evaluation window. Its proposed label rewards stocks that gain more, experience smaller drawdowns, and break out earlier. A model trained to predict this…
This research summary proposes that trading activity can reveal information about informed traders. It reports a positive relationship between turnover and next-day share-price volatility: unusually active trading may precede either a sharp rise or a sharp…
The document explains a limitation in a Chinese equities platform’s daily backtest matching: it assumes a stock at its limit-down price cannot be sold. That assumption can miss cases where buy orders remain queued at the limit price, as in the example of a…
This forum post asks how a portfolio-sizing expression allocates cash for purchases and sales. It defines an average cash amount by dividing portfolio value by the holding period, caps buy cash at available cash and a staging-dependent multiple of that…
This short introductory note defines quantitative investing as using numerical and mathematical methods to describe market behavior, identify patterns, and inform investment decisions. It presents the approach as an attempt to make decisions more objective…
The document discusses a backtesting problem in a traditional trading strategy: filtering the data to include only instruments with a positive value for a buy signal improves reported performance, but also changes the sample available to the sell condition.…
This note groups four risk-based portfolio approaches—global minimum variance, equal risk contribution, maximum diversification, and equal weight—within one optimization framework. The framework uses three adjustable parameters to represent the individual…