Replication of Past Portfolio Outperformance: Dutiful Quant Spins Cautionary Tale With Resolution
28 Pages Posted: 8 Feb 2023
Date Written: February 5, 2023
Abstract
Granted, it is generally understood that using past performance as the basis for either choosing an investment alternative or for selecting a particular configuration of a given alternative poses the risk of mistaking chance outperformance for something that might endure. But if many of the alternatives and configurations don't actually have performance staying power then the greater the number under consideration, the greater the risk of settling upon a chance outperformer. And a great deal of attention needs to be paid to that. Herein a unified approach is presented (with open-sourced code) that combines in-sample and out-of-sample testing methods in such a way as to comprehensively succeed in resolving and improving the true prospects for replication of past performance. Portfolios are formed that are optimized composites of prospective investment alternatives, that by design have fully known and replicable out-of-sample performance (with statistical significance accurately assessed).
Keywords: replication, p-value, cross-validation, multiple hypotheses, FDR
JEL Classification: C120, C460, C150, C870
Suggested Citation: Suggested Citation