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.
21,023 documents
This research summary explains a model in which product-market competition and expected asset returns influence one another. Competition can reduce profit margins, increasing firms’ operating leverage and exposure to systematic shocks. Threats from new…
This Chinese-language research summary describes a top-down framework for institutional fund allocation that connects macro asset views with strategy and product selection. It argues that asset allocation recommendations need to flow through to specific…
The document defines a Hurst oscillator using two smoothing steps. First, it calculates a centered moving average by offsetting a simple moving average of median price by half the selected period. Separately, it assigns a flow value based on the close’s…
This summary of a May 2020 China A-share market-timing report combines sector performance, valuation, market breadth, fund positioning, and an intermediate-term model. It lists the strongest and weakest sectors, notes that valuations fell across several…
The document reviews research on whether crowded mutual fund ownership predicts subsequent stock returns. It defines stock-level crowding as active mutual fund ownership divided by average share turnover, using lagged holdings and turnover data. Stocks are…
This Chinese-language article outlines a stock-selection screen that looks for shares with daily amplitude above 1%, excludes stocks that hit the daily limit on the previous session, and uses a morning star candlestick pattern as a further filter. It frames…
This tutorial explains how convolutional neural networks can improve on a simpler model for classifying MNIST handwritten digits. It reports accuracy of about 99.2% for the CNN, compared with 91% for the earlier model, but gives no experimental setup or…
This document explains introsort, a comparison sort that begins with quicksort and switches algorithms to control performance. It uses heapsort when recursion depth risks becoming excessive and insertion sort for small partitions, keeping quicksort for…
The Phase Change Index (PCI) is presented as an indicator for distinguishing consolidation, uptrend, and downtrend phases. It compares price deviations with a linear path over a lookback period, then scales the positive and negative deviations into an…
This document describes a stock selection filter using amplitude above 1, a positive price-to-earnings ratio, and a turnover rate between 3% and 12%. It frames these as measures of price movement, profitability or valuation, and trading liquidity. The…
This research summary reports tests of alternative and growth factors for selecting Hong Kong stocks. Of 17 alternative factors examined, most showed relatively weak selection effects; realized return kurtosis over 240 days and the ratio of recent to…
This 2018 research summary assesses the outlook for Chinese equities in 2019 using earnings, credit growth, producer prices, retail sales, and valuation. It argues that slower growth and gradual valuation improvement limit broad market upside, and advises…
The post explains how to estimate the return over the first hour of a trading day when the available price data does not include the 9:30 opening price. It uses the prior session’s 15:00 close as a substitute starting price and compares that value with the…
The article uses gambling and trading stories to explain why survival and capital preservation come before pursuing returns. Roulette illustrates how a small house edge compounds with repeated play, while blackjack card counting shows how a positive edge can…
This indicator overlays the price history of two currencies or other assets on one chart so a trader can inspect their relative movement. The example describes viewing USDCHF alongside EURUSD and presents the tool as potentially useful when studying…
This implementation describes a systematic filter for candidate equity pairs or larger baskets. It first constructs a spread using a chosen hedge-ratio method, including ordinary or total least squares, minimum half-life, minimum ADF, Johansen, or Box–Tiao…
The document explains the Doji candlestick pattern and describes an automated detector that scans charts for it. A candle qualifies when its opening and closing prices are close together, with those prices near the midpoint between the candle’s high and low.…
This post raises questions about the usefulness and timing of Chinese stock opening-auction data. It proposes deriving unmatched order volume by subtracting cancellation and matched-trade volume from total submitted volume, then estimating an unmatched…
The discussion addresses a question about why a strategy cannot retrieve current factor values. The reply suggests the issue may be that some stocks or dates produce missing values, rather than that factor retrieval has failed altogether. It recommends…
This forum post investigates a simulated-trading run that fails while rendering performance charts with BigTrader v9. The author suspects that the newer module cannot draw results when the run covers only one day, unlike an earlier backtesting module.…
The document explains the Factory Method design pattern, a creational approach in which a creator declares a method for producing a product interface and concrete creator subclasses choose which product implementation to instantiate. The sample uses a…
The article describes a stock-selection screen combining a turnover-rate band with beverage and alcohol import-export classification, a robotics concept label, and a maximum circulating market value. Its accompanying discussion says the robotics theme is…
This introductory guide describes a supervised machine-learning workflow for quantitative equity strategies. It begins by defining the stock universe and target, such as a future return, volatility, or ranking, then splitting historical observations by time…
This Chinese A-share screening proposal combines an amplitude above 1, a prior-day 9:15 indicative auction price at the limit down level, and 2021 revenue greater than 1.1 times 2018 revenue. Its final version also adds 2021 net profit growth above 5%. The…