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 article describes a framework that links macroeconomic and style factors with traditional asset-class allocation. It proceeds from selecting factors and estimating asset exposures to building factor-mimicking portfolios, forecasting their returns,…
The article compares long and short training windows for an AI stock selection strategy and recommends evaluating each window against a fixed validation period. In its rolling experiments, extending the sample from 2005 to 2021 changed labels, factor…
This submission outlines an intraday stock-selection idea for a day when a market theme is breaking out. It proposes identifying a popular sector early, using large orders that hold at the daily price limit as a sign of a clear direction, then ranking…
This Chinese post compiles a selection of 65 titles from a much larger collection of publicly released Springer books, focusing on data and machine learning. The bibliography spans foundations such as algebra, probability, statistics, optimization, and time…
This research summary reviews several quantitative approaches to interpreting China’s 2018 equity market. It reports historical analysis of rebounds in the CSI 1000, saying smaller-cap indexes tended to outperform larger ones during rebounds, while sectors…
This BigQuant framework describes a workflow for computing and evaluating minute-frequency stock factors. Researchers define factors in SQL against a specialized derived minute-bar table, assign an output table name, and run the program to calculate and…
This document is a brief outline of a presentation on machine learning in finance. It names four application areas: Lasso regression for commodity futures price prediction, decision trees for detecting possible financial fraud, logistic regression for…
The document introduces fund-of-funds structures and distinguishes four types according to whether the parent and underlying funds are actively or passively managed. It focuses on an actively managed parent investing in passive sector ETFs, and describes an…
This brief strategy description proposes checking stocks one by one to identify which has produced the highest return under an event strategy, then buying that stock. It specifies daily Chinese stock-bar data, a backtest beginning in 2020 and running through…
This older Chinese-equity strategy looks for stocks that rally to the daily limit, pull back, and later break to a new high. It defines a pullback as any post-limit-up close below the earlier limit-up price. After the pullback, a new high triggers a purchase…
This Chinese-language conference excerpt introduces how artificial intelligence is being adopted by global asset managers. It frames technology as one response to falling margins per unit of managed assets, alongside efforts to grow assets under management.…
The document discusses how to judge excess returns in an index enhancement strategy. It argues that alpha should be evaluated by its persistence and consistency against a benchmark, rather than by the size of gains in a brief period. It describes monthly…
This brief BigQuant support exchange concerns an error triggered after a user changed features in a beginner template. The response identifies a formatting issue in the feature list: comments or notes should be placed on separate lines rather than appended…
This document outlines a Turtle style trend following strategy for stocks. It buys when the close crosses above the prior 20 trading days’ high and exits when the close crosses below the prior 10 trading days’ low. The examples define signals using current…
The document uses national input-output tables to map intermediate flows between industries and represent those relationships as networks. It compares China across several historical snapshots with the United States in a later snapshot, using network…
The post asks how to calculate, for each instrument, the rolling five-day count of sessions when a stock value exceeds an index value, then assign those counts back to the original dataframe. The author reports that the grouped calculation produces the…
The document examines a top-ranked strategy that selects the smallest stocks in the CSI 1000 and reports unusually strong historical performance. Its central lesson is that a backtest must use index constituents as they were known at each historical date.…
This equity factor study measures buying and selling pressure through the amount of time a stock spends at different positions within its recent price range. It first normalizes price between the interval’s high and low to obtain relative price position…
This Chinese equity research note develops industry-rotation signals from several types of institutional money flow. It treats a flow source as useful when its derived signals show a reasonably orderly relationship across ranked groups and a tolerable…
This Chinese-language post introduces a custom-coded stock selection strategy on the BigQuant platform. It says the strategy combines five selection rules, mainly seeking stocks that have fallen sharply and begun to stabilize, while also attempting to…
This Chinese-language post describes a stock selection screen combining three conditions: RSI below 65, the day’s volume above 1.05 times the prior day’s volume, and the absolute move from the previous close to the opening price below 6%. It frames the…
The note recasts a past trading approach as a daily double moving average strategy, using crossovers as signals. The author reports that this approach produced losses over several consecutive years, offering a cautionary personal outcome rather than a…
This short Chinese-language Q&A explains the delayed-entry setting in factor analysis. The setting shifts the point at which stocks are purchased based on factor values by a specified number of days. The question concerns a one-day rebalance cycle and asks…
The document describes adding a custom Rank Information Coefficient module to a stock-ranking strategy. The module reports average RankIC separately for the training set and the test set, providing a way to assess how well a model’s rankings align with…