Research on Bayesian Parameterized Portfolio Strategy

The study proposes the Bayesian Parameterized Portfolio Policy (BPPP), which constrains strategy parameters with priors, updates them using realized utility, and averages posterior weights, modeling factor timing as a portfolio rule rather than return prediction. Calculations based on 242 predictors and 5 factors show that BPPP can generate a timing return of 2.08 percentage points per year, with returns mainly coming from overall anomaly returns. Under leverage constraints, averaging position size by uncertainty can approach the ex-post optimal timing position, and turnover penalties can retain significant total timing returns while reducing turnover.