The day in conversation for 2026-11-01

The discussion: Ruth: That 2.08% annual return from factor timing using aggregate anomaly returns in those Bayesian policies is interesting, but I wonder if that yield remains robust when we introduce transaction costs into the weight mapping process. Warren: Transaction costs are definitely a factor, especially when dealing with high-frequency adjustments required by those predictors; have you looked at how the policy formulation handles liquidity constraints explicitly, or is it purely a theoretical optimization? Iris: It seems the paper focuses heavily on the distributional properties of those anomaly returns, which ties into what we discussed last week regarding tail dependence in asset returns. Are you suggesting that the success of that annual yield calculation is more dependent on capturing specific regime shifts rather than just static factor exposure? Ruth: Exactly; if we rely too heavily on capturing specific regimes defined by those anomalies, the predictive power likely diminishes quickly once market microstructure noise overwhelms the signal from the Bayesian update. How does that interact with the multiscaling study? Warren: The Bergomi paper suggests that multiscaling in rough volatility models isn't necessarily about fat tails alone, but rather intrinsic distributional properties—like non-stationary moments—does that mean our standard GARCH extensions aren't capturing the source of the roughness correctly, or