Testing Handcrafted Portfolio Weights Against Common Alternatives
Summary
This article evaluates a hierarchical, handcrafted method for allocating weights among trading rules and instruments. The method is intended to be transparent enough for spreadsheet implementation, grounded in theory, and responsive to uncertainty in estimates. The test portfolio includes carry and several exponentially weighted moving average crossover speeds. The author compares the handcrafted weights with naive Markowitz optimization, bootstrapping, shrinkage, and equal weighting, examining both forecast weights and instrument weights.
The displayed outputs show the handcrafted method organizing rules into intuitive groups, such as faster and slower momentum plus carry, and instrument allocations across asset classes. The author characterizes naive Markowitz weights as unstable, while reporting that handcrafted and equal instrument weights are relatively similar because the allocation is broadly spread. For forecast weights, the tested methods allocate a larger share to carry than equal weighting does, though the difference varies by instrument. The post concludes that the method produces robust weights and decent out of sample performance, but the excerpt supplies limited numerical comparison of performance and does not detail the full test design, assumptions, or uncertainty around those conclusions.
Key ideas
- The handcrafted method organizes allocations hierarchically and aims to make weight choices transparent and practical.
- The comparison includes naive Markowitz, bootstrapping, shrinkage, and equal weighting.
- The displayed rule hierarchy groups faster momentum, slower momentum, and carry separately.
- The author reports that naive Markowitz weights vary substantially, while handcrafted instrument weights resemble equal weights more closely.
- The excerpt asserts robust weights and out of sample performance but offers limited detail on the evidence behind those claims.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.