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.
5,922 documents
This literature summary covers two separate topics. The first describes a framework for studying how mutual fund managers’ personal characteristics relate to performance, decomposing outcomes into overall results, return and risk, and timing and security…
This document summarizes an ETF tracking workbook and describes the market landscape it was designed to monitor. It reports that broad-market and sector or theme ETFs formed the largest non-money-market categories, with examples of major index exposures and…
This brief strategy description proposes following shareholdings associated with Hong Kong Securities Clearing (Nominees), China’s social security funds, and qualified foreign institutional investors. It identifies these entities as the focus of a…
This reference explains how an engine represents cash, margin, betting, and wallet accounts across backtests and live trading. It defines total, locked, and free balances and requires them to reconcile at currency precision. It also describes how pending…
This article describes 25 reasons retail investors may lose money in equities. It groups them around information and research gaps, overreliance on one strategy, emotional and herd-driven decisions, limited planning and experience, concentrated portfolios,…
This review explains why diversification may help more during declines than during gains, and why average correlations can hide that difference. It recommends estimating an asset pair’s downside relationship by conditioning on the portfolio’s main growth…
This review challenges the assumption that diversification works equally in rising and falling markets. It argues that the useful question is how a prospective diversifier behaves when a portfolio’s main growth asset falls. To measure downside correlation,…
This review describes a method for identifying crowded markets and distinguishing potential bubble build-up from unwinding. It measures sector or factor clustering from return covariances and principal components, then pairs that measure with relative…
This configuration specifies a Qlib time-series Transformer workflow for predicting Chinese CSI 300 stock returns from Alpha158 features. It defines a Chinese market data source, uses the CSI 300 as the instrument universe and SH000300 as the benchmark, and…
The tutorial explains how Brinson performance attribution separates a portfolio’s actual return from its benchmark return and decomposes active performance into timing, security selection, and an interaction component. The benchmark reflects planned asset…
The study summary describes a dynamic contextual alpha model designed to address a weakness in conventional multi-factor scoring: applying the same factor evaluation across the entire market. It groups A-share stocks by characteristics such as size,…
This example combines three cross-sectional stock measures into a composite score: market capitalization rank, turnover rank, and the rank of the latest close-to-prior-close ratio. It filters out risk-warning stocks, recently listed stocks, certain listing…
This weekly competition report profiles the ten highest-ranked ModelScore factors and tracks a CSI 1000 index-enhancement portfolio. It summarizes average style exposures from daily multivariate regressions and reports average within-industry Rank ICs, the…
This discussion examines how growth in assets can affect a quantitative strategy’s returns and stability. It describes a trade-off: as a strategy takes more capital, returns may decline and volatility may rise. It distinguishes broad-capacity equity…
This study examines how neural network hidden-layer configurations affect reinforcement-learning performance in stock portfolio optimization. Using the FinRL framework, agents receive open, high, low, close, and volume data and seek to improve portfolio…
The article describes using a deep neural network to estimate risk for an automated, beta-neutral long–short equity strategy. Financial statement data and historical prices serve as inputs; the model predicts return quantiles over horizons from five to 90…
This paper summary presents a method for evaluating active mutual funds when their stated benchmark may not match their actual style. It argues that omitted risk factors can make factor-model alpha estimates too high, while benchmark mismatch can also…
This report describes a regularized fund-evaluation model used to rank Chinese equity funds and form fund-of-funds portfolios. It compares funds selected for overall ability, stock-selection ability, and market-timing ability, then reviews their holdings and…
This explanation clarifies why a stock strategy may send orders for shares it already owns. It separates current holdings into positions with sell signals, new buy candidates, and existing positions that still qualify to be held. The strategy sells the first…
This article explains how to allocate weights in a multi-factor equity portfolio using a structured risk model. It decomposes stock returns into factor exposures, factor returns, and stock-specific returns, then describes industry and style neutrality…
This report summary outlines a risk-budgeting approach to domestic multi-asset allocation conditioned on economic growth and inflation regimes. It describes statistical comparisons of asset behavior across those states and reports rotation patterns: bonds…
This research note describes a China A-share fund designed to enhance returns over the CSI 500 while managing benchmark-relative risk. It attributes the strategy’s excess returns to quantitative multifactor stock selection, sector allocation, trading…
This July 2020 market commentary assesses whether a brief cooling in China’s hot A-share market signaled an end to the rally. It frames the issue as a balance between incoming capital and potential selling, identifying new fund launches, northbound foreign…
The discussion addresses how to search for effective weights when combining two stock-ranking factors. It proposes using BigQuant’s parallel tuning module, m.tune, rather than treating each weight as a separately exposed parameter. The key implementation…