Pre-Mainshock Occupancy Entropy × WCM

We report a negative result and, more importantly, the audit trail that produced it. Within the WCMpolytope-occupancy framework we tested whether the magnitude-class occupancy entropy ofmicroseismicity rises before large shallow mainshocks in the Japan Meteorological Agency (JMA) catalogue(1998–2023). An initial discovery on the 2016 Kumamoto sequence showed a strong positive signal. A pre-registered frozen hold-out (v1.3; eight eligible targets) did not confirm it (cohort mean standardized rankeffect T = +0.467, cohort outer-null p = 0.194), and post-hoc inspection showed the positive contributionconcentrated in the 2011 Tohoku sequence. We then built a larger, mechanically defined stress test (DesignC): starting from 131 shallow (depth ≤ 60 km) M ≥ 6.5 mainshocks nationwide, we mechanically excludedtargets within ±90 days and 5° of any M ≥ 8 event and required a valid pre-window, yielding 54 effect-blinded, pre-frozen targets. The predicted high-entropy direction did not reappear (T = −0.136, p = 0.857).The negative conclusion was robust to two implementation discrepancies discovered during audit: a circular-versus-rectangular geometry mismatch (circular re-analysis T = −0.103, p = 0.770) and a 53-versus-54 cohort-size asymmetry from one non-estimable target (fully symmetrized re-analysis T = −0.136, p = 0.852). Amethodological by-product is that an event-centred analysis pipeline can induce a non-zero empiricalbaseline even when its component rank statistic is theoretically centred, so pipeline-level empiricalcalibration is required. We conclude that a general high-entropy pre-mainshock occupancy precursor is notsupported under the tested framework, and we document the full contraction of the claim as the primarycontribution. ※ The license for codes is AGPL3.0-or-later