Ze-AL v4 Stage-0: Activity-only Bayesian-surprise allostatic load (pre-registered simulation)

Stage-0 simulation testing six estimators of allostatic burden across nine sensitivity conditions and three outcome targets, with pre-registered decision rules locked before any production run. Tests whether replacing the original v3 σ_E(t) normalisation with a frozen calibration variance (σ_calib) plus multivariate Mahalanobis distance and a Bayesian KL-surprise term resolves the self-cancellation problem reported in v3 and outperforms a 24-hour moving-average residual baseline. Pre-registered seed = 20260509. Code: https://github.com/djabbat/Ze-AL-v4-private (private; read-only access on request). Stage-1 (real-data validation in NHANES, UK Biobank, and Whitehall II) is planned as a separate OSF project after Stage-0 publication. Author: Jaba Tkemaladze, Georgia Longevity Alliance. Conflict of interest: developer of the Ze-AL formalism; pre-registration and deterministic public code mitigate developer bias.