Testing Portfolio Optimization with Different Clustering Choices
Summary
The document describes an experiment comparing clustered and unclustered portfolio optimization across trading rules and instruments. It varies in-sample and out-of-sample periods, the number of assets, correlation shrinkage, Sharpe ratio shrinkage, and the number of clusters. Randomly selected configurations are evaluated repeatedly, with forecast weights examined across combinations of sample length and portfolio size.
The available text reports only a few qualitative observations: not clustering and excessive Sharpe ratio shrinkage appear unfavorable in some short-sample cases, while several other configurations show no significant downside. It provides no charts, numerical results, final summary, or enough detail to reproduce the experiment. The findings should therefore be treated as provisional; the document itself emphasizes that significance is limited in many cases, and does not establish one universally best cluster count or shrinkage setting.
Key ideas
- The experiment varies cluster count, correlation shrinkage, and Sharpe ratio shrinkage in portfolio optimization.
- It compares results across different asset counts and in-sample and out-of-sample periods.
- Some short-sample combinations appear to penalize no clustering or excessive Sharpe ratio shrinkage.
- Many tested combinations show no statistically significant weakness in the text provided.
- Missing charts and detailed numerical results limit interpretation and reproducibility.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.