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 describes a macro allocation framework built from six risk factors covering growth, inflation, rates, credit, foreign exchange, and term spreads. It combines momentum signals with cycle-phase estimates, maps the resulting states into…
This research summary compares monthly factors built from price and volume data with quarterly factors built from financial statements. It reports that market data factors were more often effective in long-short portfolios, and examines their relationship…
This post describes a Chinese equity screen combining a daily amplitude threshold, exclusion of stocks that closed at a limit-up the prior day, and a positive MACD reading. The stated rationale is to favor stocks with room to move that also show positive…
The report outlines a multi-asset allocation framework that links macroeconomic risks to broad asset classes, styles, and industries. It identifies economic growth, interest rates, exchange rates, inflation, and credit as five central risk factors. A…
The research note studies tail dependence between individual stocks and the broader market. Using copulas, it estimates upper-tail and lower-tail correlation, which describe whether assets tend to move together during extreme outcomes. The reported…
The discussion explains a rolling sequence window as a way to assemble recent observations into a single training example. For instance, a model can receive a window of past factor values spanning several days, with the window advancing through time to…
This tutorial introduces recurrent neural networks as models for sequential data, then explains long short-term memory networks as a design intended to address the difficulty ordinary RNNs have retaining information across long sequences. It describes the…
This strategy description applies a simple price rule to one Chinese stock. It defines a reference high as the highest closing price over the past 20 days and a reference low as the lowest closing price over the past 10 days. The stated entry condition is…
Average True Range (ATR), developed by J. Welles Wilder, measures price volatility rather than direction. The document defines true range as the greatest of the current high–low range and the distances from the previous close to the current high or low. ATR…
This Chinese equity screen selects stocks with amplitude above 1 and previous-day turnover above 8%, while excluding Beijing-listed shares. The article presents amplitude as a way to locate more volatile stocks and turnover as a sign of recent trading…
This Chinese-language strategy note tests a simple stock screen for mainland China A-shares. It selects companies using three annual-report criteria: net profit growth above a threshold, return on equity above a threshold for three consecutive years, and a…
This introductory tutorial explains machine learning concepts and linear regression with finance examples. It distinguishes features from labels, outlines training, validation, and test data, and contrasts supervised with unsupervised learning,…
This research review applies conditional generative adversarial networks (cGANs) to multi-asset allocation. It conditions the model on recent asset-return sequences and random inputs to generate possible future paths, then estimates expected returns and…
The document raises a position-management idea for an equity holding: after a stock gains 10%, sell half the shares to secure some of the unrealized profit while leaving the rest invested for further gains. Its example starts with 1,000 shares and proposes…
This research summary describes two strategies using constituents of the Shanghai 50. The momentum version ranks stocks by their return over the previous three trading days, selects the three strongest, weights them equally, and holds each cohort for nine…
This brief Chinese-language entry describes a lecture on deep reinforcement learning by researcher Matteo Hessel. It introduces general value functions (GVFs) as auxiliary tasks and mentions methods for handling scaling problems in reinforcement learning…
This weekly report compares stock selection portfolios built with several artificial intelligence methods across broad Chinese A-share and index-constituent universes. It separates results for unconstrained selection, industry-neutral portfolios, and…
This report evaluates momentum based sector allocation in Chinese equities. It compares ordinary lookback return ranking with risk-adjusted alternatives, including Sharpe ratio, information ratio, and a path-adjusted measure that divides return by the sum of…
This report summarizes quantitative approaches to enhancing returns over Chinese equity benchmarks while managing tracking error. It frames enhancement as a blend of active and passive investing and describes a top-down range of choices: market exposure,…
This document outlines a monthly long-only stock selection strategy for the Chinese A-share market. It ranks the universe using a market-cap rank and a 20-day average of price changes, combines the two into a score, then holds 50 equally weighted stocks with…
This research note explores additional technical factors for stock selection in the CSI 500 universe. It groups candidate measures into turnover behavior, illiquidity, and combined volume-price information. Turnover candidates include stability, coefficient…
This report describes an initial step in building a multi-factor equity model: choosing which factors to combine. It proposes reducing the factor search space by including exposure relative to a benchmark as one screening criterion, alongside the strength of…
This example strategy searches Chinese A-shares for stocks that recently had a long run of consecutive limit-up sessions and then broke the streak. It further requires positive average returns over the three days after the break, excludes several board…
The page explains a practical issue in BigTrader’s simulated-trading workflow: strategy calculations run in a container on a daily schedule, so in-memory values may not carry over between runs. If a strategy needs to retain information such as its net asset…