Compare the Input Requirements of ACSA and KSA

Determine which input requirement is more demanding in practice: the ACSA requirement for a quantile predictor of the conditional portfolio gain, or the KSA requirement for a centering predictor of the scenario gain.

Background

Adaptive Conformal Scenario Analysis (ACSA) requires a predictor of the conditional portfolio-gain quantiles, whereas Kernel Scenario Analysis (KSA) requires a centering predictor representing the predictable component of the portfolio gain. The paper notes that the relative difficulty of these requirements is not established. A conditional median may be less sensitive to outliers and easier to estimate than a conditional mean, while extreme quantiles may be difficult to estimate because they are supported by relatively few observations. Resolving which requirement is more demanding would clarify when each method is preferable and how their practical data and modeling burdens compare.