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 documentation explains how to define model inputs and prediction targets in FreqAI strategies. It distinguishes base features, such as price indicators, volume, and time variables, from configuration-driven expansions across periods, timeframes, shifted…
The document defines a DMA indicator using an applied price and a period measured in days. It gives a default chart timeframe of M15 and a default period of 17. Its recurrence updates the current value as a weighted blend of the applied price and the…
This ProRealTime version 11 indicator builds a historical distribution of price changes by comparing each close with the price from a configurable number of bars earlier. It separates gains and losses, rounds percentage moves into buckets of 0.1 percentage…
The document describes an indicator based on the Balance of Market Power concept. Its version smooths values with a Jurik filter rather than a simple moving average and constructs bands using a Jurik-based deviation measure instead of standard deviation. The…
This Chinese-language forum post addresses two data needs in quantitative stock research: constructing weekly bars from daily observations and obtaining historical quarterly financial measures such as return on equity. It outlines grouping daily data by…
This report summary describes a fund-selection approach that uses portfolio holdings to estimate managers’ stock-picking and market-timing abilities. Holdings provide information for attributing and forecasting performance that is not available from fund net…
The document presents a framework for testing stock-selection factors. It evaluates each trading day’s stock factor values against returns over subsequent holding periods, aiming to show how predictive information changes as the holding period grows. Daily…
The post reports an observation from more than six months of intraday modeling in Chinese equities using five-minute bars. At each time point, the author independently tuned parameters with tree-structured Parzen estimation. The reported optimization…
This research summary examines where mutual fund excess returns may come from, separating security-selection skill from timing skill. Timing is framed as adjusting a fund's market beta, with stronger exposure during rising markets and reduced exposure during…
The document introduces quantitative investing as an approach that uses numerical models to build an investment process, contrasting it with methods that rely more on qualitative judgment. It describes how quantitative methods can combine computing with…
The document describes an asymmetrical RSI indicator attributed to Sylvain Vervoort. It accepts a calculation period, applied price, and overbought and oversold thresholds. Price changes are computed as differences from the previous applied price, then…
This document describes a price channel built from centered linear regressions of high and low prices over a configurable lookback, set to 300 periods by default. It calculates separate regression lines for highs and lows, then offsets each line by the…
This short assignment response defines quantitative investing as turning recurring patterns observed through statistical analysis of financial markets into structured methods for future trading. It mentions stocks, futures, and digital assets, and describes…
This stock-selection example combines a turnover-rate band of 3% to 12%, the sign of daily price change multiplied by large-order net flow, and a MACD condition. The stated interpretation is that the turnover filter selects a desired level of activity, while…
This guide organizes U.S. equity research inputs into five source groups: company filings, investor relations materials, government macroeconomic releases, trading calendars and halt notices, and public checks on brokers or advisers. It treats regulatory…
The document defines a sine-weighted moving average that smooths an applied price over a configurable period. Each price in the lookback window receives a weight based on a sine function: weights rise across the window toward the most recent observation, and…
This indicator compares average applied prices from bullish and bearish candlesticks over a chosen period. It calculates a separate mean for each direction, then displays their difference. The result is positive when the bullish-candle average is higher and…
This post presents a Chinese-equity screening rule combining daily price amplitude above 1%, a K-line indicator below 20, and relative volume greater than 1.5 but less than 6. It offers formula and Python-style examples for calculating amplitude from the…
ADXm is described as a variation on the traditional ADX indicator that makes trend direction easier to read. It displays positive and negative ADX half-waves separately and shows the difference between the positive and negative directional indicators. Dashed…
This equity screen combines three filters: daily high-to-low range above 1%, a reversal or engulfing-style pattern, and a concentration measure below 20%. Formula and Python examples illustrate how the conditions might be combined to produce a candidate…
This indicator presents a histogram intended to compare inferred activity by investors and speculators. It computes delta as the investor measure minus the speculator measure, then assigns each bar’s accumulation/distribution value to one side according to…
The Average Range indicator sets reference levels around the opening price of a selected period. It describes four bands: two at half the average range above and below the period open, and two at the full average range above and below it. The period can be…
This weekly market report organizes several indicators of Chinese equity liquidity by sector: ETF subscriptions and redemptions, northbound investor flows, estimated allocations from pending and newly raised equity funds, company buybacks, and…
The document explains a two-stage model that uses gradient-boosted decision trees to create features for logistic regression. The trees split and combine the original inputs; each sample is then represented by a binary vector marking the leaf it reaches in…