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.
16,761 documents
This weekly report compares artificial intelligence stock-selection portfolios across broad A-share universes, industry-neutral portfolios, and portfolios restricted to the CSI 300 or CSI 500. It names XGBoost, support vector machines, random forests,…
The document presents a Chinese A-share backtest question: after selling a holding at the open, a daily-frequency strategy places a purchase near the close but buys only a small amount. The author wonders whether the simulator updates available cash only on…
MAVA-XonaX describes a moving-average crossover approach for EURUSD on a four-hour chart. It uses four six-period moving averages calculated from bar open, close, high, and low prices. A cross of the close-based average above the open-based average is…
This note proposes screening Chinese stocks listed in 2021 for daily turnover between 3% and 12%, then keeping those whose opening price is within five percent of the 10-day moving average. The approach combines a trading-activity range, an IPO-year cohort,…
This note outlines a Chinese equity screen using three conditions: positive but moderate returns over the prior ten days, appearance on the previous day’s exchange trading-activity list, and descending ranking by a measure called fund strength. The intended…
This note describes a Chinese equity screen that starts with stocks in the metaverse theme, filters for prices above the 250-day moving average, then ranks candidates by the day’s auction amount. It proposes selecting the top five and also mentions ordering…
The study defines an earnings surprise by comparing reported results from forecasts, flash reports, or periodic filings with analyst consensus net profit estimates from the day before the announcement. For nonannual reports, it estimates full-period…
The document explains lookahead bias: a backtest can accidentally use future candle data because the full historical dataframe is loaded before indicators and signals are calculated. This can make results appear unrealistically strong. It describes an…
This Chinese-language forum exchange discusses excluding special-treatment, suspended, and delisted shares from a stock universe. A reply provides two approaches based on historical name-change records: retrieve securities available around the relevant date,…
This EUR/USD strategy is described for a 30-minute chart and aims to participate in both short ranges and larger trends. It opens buy or sell positions when a one-period smoothed moving average crosses the 200-period smoothed average. It then uses crossovers…
This five-minute long-only strategy combines a fast stochastic oscillator, an ADX strength filter, and short EMAs. It enters when the opening price is below the five-period EMA of lows, the stochastic %K crosses above %D while both remain below a tunable…
The document describes a ProBuilder implementation of Larry Williams’s Ultimate Oscillator for use in backtests and automated trading systems. It calculates buying pressure relative to true range over three lookback windows: a base period, twice that period,…
This Chinese-language report summary reviews the history and practice of machine learning in quantitative investing. It says that machine learning was used in the field during the early-1990s boom and that its applications continued in algorithmic trading…
This report overview describes the longstanding use of machine learning and artificial intelligence in quantitative investing. It notes that applications were already present during an early-1990s wave of interest, and that use continued in areas such as…
This Gold trading system turns an observed relationship between price waves and the stochastic oscillator into entry and exit rules. It tracks stochastic values relative to upper and lower thresholds, then looks for a crossover against a smoothed stochastic…
This post proposes a rules-based screen for Chinese A-share stocks in a designated metaverse sector. It combines a recent trading-activity condition with a profitability filter requiring return on equity above a stated threshold across five consecutive…
This study develops two equity factors from the relationship between intraday volatility and the return-to-volatility ratio. The “rebuilding” factor uses the daily covariance between that ratio and volatility to represent cases where volatility rises without…
This weekly report compares machine-learning stock-selection portfolios across several Chinese equity universes: broad-market stocks without industry neutralization, broad-market selections neutralized to the CSI 300 or CSI 500 industries, and selections…
This document outlines an expert advisor intended for XAUUSD on a 15-minute chart. It combines RSI entry filtering with recovery entries spaced according to ATR, plus market-structure checks based on lower lows, lower highs, and support conditions. Position…
This proposed A-share stock screen filters for companies with turnover between 3% and 12%, market capitalization below 10 billion yuan, and no losses, then selects stocks with daily returns from −5% to 2.6%. The accompanying formula and Python example…
This configuration defines a Qlib workflow for training a TabNet model on Alpha158 features for CSI 300 stocks, using Chinese market data and the CSI 300 index as benchmark. It sets a historical data window, separates fitting, validation, and test periods,…
The Pollan indicator combines the Commodity Channel Index and Relative Strength Index by comparing their values across a fixed window of historical bars. It forms two oppositely oriented series from the CCI–RSI and RSI–CCI differences, with a coefficient…
The post proposes a stock screen combining a daily high-low range threshold, rising price lows, and return on equity above 15% for five consecutive years. Its rationale is to pair a technical sign of improving price support with a record of strong…
The document describes a tool for checking whether a trader’s actions and recent outcomes are associated with subsequent losses. It trains a neural network on eight features from closed trades, then reports validation accuracy relative to a majority-class…