Skip to content
All library documents

Accounting for Trading Costs in Portfolio and Forecast Weight Optimization

Article Systematic trading blog (Rob Carver)

Summary

This document compares ways to include trading costs when optimizing portfolio or forecast weights. Options include optimizing gross returns, subtracting costs to form net returns, optimizing costs alone, penalizing costs by a multiplier, applying a maximum cost threshold, or calculating weights from gross returns and adjusting them afterward. It also discusses pooling information: gross returns can be pooled across instruments, while costs may remain instrument-specific, or both can be pooled. The examples contrast cheaper and more expensive markets and show how cost treatment affects forecast allocations.

The author’s preferred approach emphasizes pooling gross return data, retaining instrument-specific cost estimates, filtering out excessively expensive trading rules, and applying a post-optimization cost adjustment. The document cautions that shrinkage on net-return Sharpe ratios may reduce the influence of costs, and that shrinkage settings depend on the use case. Example allocations and configuration choices illustrate these points, but the excerpt does not provide a controlled performance comparison across methods. Results therefore support practical guidance rather than a universal ranking; cost estimates, pooling choices, and optimizer assumptions remain important limitations.

Key ideas

  • Ignoring costs can favor high-return assets whose trading expenses are also high.
  • Costs can enter optimization through net returns, a penalty multiplier, thresholds, or an adjustment after gross-return weights are calculated.
  • Pooling gross return data across instruments can help, while instrument-specific cost estimates preserve differences between markets.
  • A maximum cost threshold can remove rules that are too expensive to trade.
  • Shrinkage settings and cost treatment interact, so the appropriate configuration depends on the data and optimization method.

Tags

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