README.md

In jluby/portfoliowalkr: Characteristic-Matched Random Portfolio Creation

Characteristic-Matched Random Benchmarks for Equity Portfolios

Portfoliowalkr is an r package which facilitates the creation of characteristic-matched random portfolios (CRPs). Following the ideas of Burns (2004) with additional constraints for passive characteristic allocations, we propose that randomly sampling from the space of characteristic-matching portfolios provides the best metric of manager stock-picking ability when all matched characteristics are categorical.

This package serves as a wrapper for the walkr function, randomly sampling the non-negative convex polytope represented by the complete solution space to Ax = b and the N-simplex. Because each portfolio in this space has no allocation bias, sampling from this set provides a representative set of characteristic-controlled counterfactual investments. These CRPs serve as a fair benchmark for portfolio performance assessment.

Because CRPs are matched on the portfolio level but the effects of continuous characteristics are not necessarily linear, this protocol is only recommended when all matched characteristics are categorical. Should any matched characteristic be continuous, benchmarks must be considered on the basis of their nearest-available matches as described by Kane & Enos (2010) and operationalized in the portfolio package. The reasoning behind this distinction is explained further in this package's documentation.

jluby/portfoliowalkr documentation built on April 4, 2020, 1:46 a.m.