Quantitative Strategy Design, Factor Processing, and Risk Management Q&A
Summary
This Chinese Q&A collection covers practical topics in quantitative stock research and strategy deployment. It explains cross-sectional versus time-series ranking, preprocessing such as missing-value and outlier handling, neutralization and normalization, and how factor signals may be used to rank stocks. It also discusses rolling machine-learning models, factor and style rotation, style exposure attribution, and using recent style similarity to adjust training-window length.
The answers outline broader portfolio practices: diversify across strategies and exposures, constrain factor or sector concentrations, allocate by inverse volatility or portfolio optimization, and combine stock selection with timing. They distinguish factor returns, asset-allocation returns, and CTA returns as different sources of performance. Practical notes address execution timing in a bar-based backtest, prediction data configuration for simulated trading, and saving model parameters. The document is an informal collection, not a tested strategy specification: many questions receive no substantive answer, some replies defer details to later sessions, and it reports no systematic performance evidence.
Key ideas
- Cross-sectional ranking compares securities within a date, while time-series ranking evaluates observations over time.
- Factor preprocessing may include missing-value treatment, outlier handling, neutralization, and normalization.
- A style-rotation framework can choose among style-oriented strategies using recent factor exposures and style similarity.
- Diversification and exposure limits are presented as core ways to manage portfolio risk.
- Inverse-volatility weighting and constrained portfolio optimization are two suggested allocation approaches.
- Backtests should reflect the simulator’s order and fill timing, which may occur on different bars.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.