Pooling Performance Estimates Across Instruments for Portfolio Weights

Summary

The document proposes selectively pooling return histories across instruments when their estimated Sharpe ratio profiles across trading rules appear sufficiently similar. It describes a clustering procedure: estimate each instrument’s Sharpe ratios, measure pairwise distances, compare each distance with a critical value based on the shorter history, and repeatedly combine the closest qualifying pair. This creates a pseudo-instrument whose returns can be compared with those of the remaining instruments.

The post situates the method among alternatives such as pooling by asset class, return distributions, or fitted portfolio weights. It reports backtest comparisons across different in-sample and out-of-sample horizons, but the table formatting is damaged and the displayed results are difficult to interpret consistently. Its closing recommendation uses pooled returns for short histories, individual returns for long histories, and a blend in between. The final weighting formula is cut off, and the document does not establish that the approach generalizes beyond its experiments.

Key ideas

• Pooling can be based on similarity between instruments’ estimated Sharpe ratio vectors across rules.
• The proposed algorithm repeatedly merges the closest pair whose distance falls below a history-dependent critical value.
• The author compares selective pooling with using pooled or individual instrument histories in backtests.
• The suggested approach favors pooled returns for short histories and individual returns for long histories.
• The final rule for blending returns at intermediate history lengths is incomplete in the supplied text.

Tags

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