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 article introduces Transformer attention and adapts the architecture to stock-factor time series. Its proposed model adds Time2Vec representations to factor inputs so the network can represent both periodic and non-periodic time patterns. Self-attention…
This talk describes how a Chinese quantitative investment firm approaches team design, hiring, researcher autonomy, collaboration, and leadership. Its central claim is that a research organization should support strategy development with data, computing, and…
The document is a brief summary of a research report on constructing a manager-of-managers portfolio. It names three stages: selecting an asset-allocation approach, evaluating investment managers, and managing the portfolio dynamically over time. It also…
This example shows a basic workflow for comparing portfolio value paths across several strategy weights. It sets a date range and starting value, generates daily returns from a normal distribution whose mean scales with the selected weight, compounds those…
This analysis examines weakening one-month return reversal and specificity factors in Chinese equities. It argues that judging a long-short portfolio alone can conceal deterioration: although aggregate reversal returns remained positive, the long leg had…
This configuration lays out a Qlib workflow for training a Transformer model on Alpha360 features and ranking CSI 300 stocks. It specifies Chinese market data, a close-to-close forward return label, robust feature normalization, missing-value filling, and…
The document outlines a proposal for a U.S. Bitcoin reserve alongside a broader stockpile of other digital assets. It says the stockpile would initially contain assets forfeited to the government, while additions to the Bitcoin reserve could use…
This research summary compares machine-learning stock-selection models trained on conventional return labels with models trained on alternative labels based on information ratio and Calmar ratio. The rationale is that these labels capture more of the path…
This research review examines whether education ties among Chinese hedge fund managers are associated with fund performance. It builds time-varying networks in which funds are connected when their managers share a school and a similar field of study, then…
The document describes a method for converting company-level factor views into industry-level indicators by aggregating the factors of each industry’s constituent stocks. It reports that mapped measures based on changes in asset turnover and leverage,…
The document describes a framework for separating active fixed-income fund returns into persistent strategic factor exposure, short-term tactical exposure, and residual return, which it treats mainly as security selection. Long-run factor betas are estimated…
This report proposes attributing bond fund returns from net asset value histories, addressing the limited holdings data available for many fixed income funds, especially private funds. Its Campisi-inspired regression model uses five risk factors representing…
This BigQuant example builds a monthly Chinese equity portfolio from a screened stock universe. It ranks eligible stocks using dividend yield and year-over-year net profit growth, then selects 50 holdings and assigns equal weights. The workflow includes…
This report summary reviews Markowitz portfolio theory and its mean-variance efficient frontier, then questions the normal-return assumption underlying variance as a complete measure of risk. Since observed asset returns can be sharply peaked and…
The roadmap describes NautilusTrader’s priorities as a production-oriented algorithmic trading engine. Its main plans are to stabilize the Rust runtime and Python bindings, improve documentation and tutorials, and make APIs and configuration easier to use.…
This report compares six ways to combine equity factors: equal weighting, historical factor returns, historical Rank IC, maximum IC, maximum IC_IR, and principal component analysis. It applies the methods within six factor groups in a Chinese A-share…
This meetup Q&A covers several practical quantitative trading topics. It explains rolling training as repeatedly updating a model using a moving or expanding historical window, and distinguishes incremental updates from retraining on the full dataset. It…
The document summarizes a research paper that uses a neural network to produce portfolio weights for exchange-traded funds. Instead of forecasting expected returns and then feeding those forecasts into a conventional optimizer, the approach directly tunes…
This research summary tests valuation factors for equity industry allocation. It examines five conventional measures, including earnings, book, sales, and operating cash flow multiples, and reports that standalone valuation rankings struggled to beat an…
This report proposes diversifying portfolios across market, style, and strategy risk premia rather than allocating only among broad asset classes. It uses representative global equity, bond, and commodity indices to construct long-only strategies reflecting…
This Python utility collection summarizes portfolio positions over time. It converts position values into allocations, identifies the largest long, short, and absolute positions, and calculates maximum and median long and short concentrations. A separate…
This article explains how to assemble historical Chinese A-share index constituents and daily price data for multi-factor research in VeighNa, using the Xuntouyan data service. Its example follows the CSI 300 over a stated historical date range. It first…
The document offers a brief allocation recommendation for bond funds in a period of increasing credit risk. It advises favoring funds with higher credit scores and specifies a threshold of 2.7 or above according to the authors’ credit scoring table. The…
This Qlib documentation explains its experiment-management structure for organizing quantitative research. An experiment manager oversees experiments, each experiment groups recorders, and each recorder represents an individual run. The high-level Qlib…