Factor Sensitivity and Research Gaps in Quantitative Strategy Building

Summary

This submission reflects on a quantitative strategy test whose results changed substantially when the author replaced its default factors or broadened the stock universe to include main-board and ChiNext shares. The observation highlights two practical research concerns: model performance may depend heavily on factor selection, and results may not carry over across different investment universes. The post frames quantitative investing as a systematic implementation and evaluation of investment ideas, while noting that building a strategy effectively—especially one using AI—remains a challenge for the author.

The author lists open learning needs rather than proposing a finished method: validating AI strategies, selecting and combining factors, deriving useful signals from high-frequency data, understanding discrepancies in financial factors across data sources, and studying strategies for stocks driven by speculative trading. No algorithm details, test design, risk metrics, or numerical results are supplied, so the reported sensitivity cannot be independently assessed. The note is useful as a research checklist, but it does not establish that any factor set or strategy is robust or profitable.

Key ideas

• The author reports that tested performance was sensitive to changing the default factors.
• Expanding the stock universe was associated with weaker reported returns.
• The submission identifies model validation and factor combination as unresolved research tasks.
• It also raises questions about high-frequency and financial data, as well as speculative stocks.
• No testing details are given to verify the reported observations.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.