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
This document describes a cross-sectional equity reversal factor that separates recent returns according to trading activity. For each stock, it looks back over 20 sessions, calculates average transaction value per trade each day, ranks the sessions by that…
The report examines Chinese analyst earnings forecasts, ratings, and target prices as inputs to equity selection. It compares six ways to combine analyst estimates and finds that weighting forecasts by analysts’ past accuracy performs better than several…
The document reports a pandas error raised when a factor-analysis workflow formats the minimum date with strftime. The shown failing expression takes the minimum of a date column and immediately formats it; pandas raises the error when that value is NaT, its…
The article introduces XGBoost as a boosting ensemble and contrasts its sequential fitting of weak learners with the parallel training used in bagging. It explains residual-based improvement, complexity penalties, sample and feature subsampling, and…
This summary of a Changjiang Securities study examines whether the weak side of equity style factors can help improve an index-enhancement portfolio. It reports historical measurements from 2007 through 2019, identifying negative excess returns for…
This report summary outlines a method for inferring informed trading from minute-level data. It first applies a “smartness” measure to identify trades attributed to smart money, then aggregates their relative execution prices into sentiment factor Q. Higher…
This short assignment response describes quantitative investing as using data to check market observations and identify patterns. It frames the goal as building strategies that can be backtested and evaluated for stability over time, then using software to…
The example constructs price-based stock factors from recent daily observations. For each half-month sampling date, it retrieves a recent window of daily high, open, low, close, and average price data for the stock universe. It calculates three mean…
This research note proposes that stock price changes reflect market forces, firm-specific short-term information, longer-term fundamentals, and noise. It uses daily minute data and cross-lag regressions of returns on changes in trading volume to estimate…
This brief report summary reviews changes in Chinese public mutual fund holdings during the fourth quarter of 2017. It says fund exposure rose in several sectors, led by food and beverage and home appliances, with increases also reported for electrical…
The document explains a portfolio approach built around StockRanker, a model that ranks stocks using historical data. Each day, the strategy buys the highest-ranked candidates and sells or replaces holdings that fall toward the bottom of the ranking. The…
The post outlines a low-frequency model for timing Chinese A-share exposure and selecting trading targets. It describes using daily Wind data and a large set of predetermined spreadsheet formulas to generate bullish or bearish signals and identify an…
This article explains the claimed tendency for Chinese A-share markets to weaken on Thursdays through three proposed mechanisms: investors selling ahead of weekend cash needs under T+1 trading and restricted weekend transfers, investors reducing exposure to…
This brief BigQuant support exchange addresses a question about a backtest whose logs and apparent trade signals stop before the requested end date. The user says the run was set through the 23rd, exchange data stopped at the 17th, and the output log stopped…
This forum post raises a practical backtesting question: whether strategy data indexed by the current date can be used to trade at that same day’s open without introducing lookahead bias. The example ranks stocks using market capitalization, tradable market…
This dated Chinese equity market report combines market activity, sentiment, sector flows, concentration measures, and style-factor performance. It describes subdued overall valuations, rising consumer-sector turnover, a recovery in TMT activity, and strong…
This review examines whether controlling for quality changes the apparent size premium in equity markets and whether small-cap exposure can strengthen other factors. It compares size-factor returns with quality and related factor exposures using time-series…
The article reviews empirical research comparing commodity futures momentum based on total returns with momentum based on residual, or idiosyncratic, returns after accounting for selected systematic factors. The study forms ranked portfolios across…
This short exchange explains how to assign groups when training a learning-to-rank model on stock observations. The questioner expects each trading date to form a group, so the model ranks stocks against other stocks observed on that date, and asks where…
This brief assignment reflects on translating discretionary investment ideas into a quantitative process. The author cites assessing future demand for a company’s product, with examples involving a stablecoin business in the context of US Treasury pressures…
This Chinese-language research digest summarizes two separate studies. The first considers adding trend-following strategies to target-date funds in defined-contribution retirement plans. It argues that these funds may concentrate too much risk in equities…
This Chinese-language tutorial describes implementing a stock-ranking strategy in BigTrader by passing StockRanker predictions directly to the trading engine and coding the trade logic there. It allocates purchases among the top-ranked stocks using weights…
This study mines short-horizon price and volume factors for Chinese equities, then examines how transaction costs and rebalance frequency affect portfolio results. It uses genetic programming to generate candidate daily factors from conventional price and…
This BigQuant note explains how to change a strategy template’s initialization settings from a declining weight schedule to equal weights. The existing example assigns larger allocations to earlier stocks using weights proportional to the inverse logarithm…