The pre-registered magnitude test failed because the statistic was bounded by construction - top-decile share spreads only 1.1-1.4x between samples and cannot express what the published oncoprotein score says varies

**Pre-registered outcome B1** was that between-sample rank ordering would agree (rho >= 0.7) while the magnitude of enrichment would differ by >= 1.5-fold, reproducing the project's standing claim that ordering transfers and magnitude does not. **Measured spread of the top-decile share** - the share of a sample's total promoter signal that falls in its own top 10% of promoters [E1]: | factor | max/min across samples | |---|---:| | H3K4me3 | 1.11x | | IgG | 1.11x | | ENL | 1.14x | | KMT2A-N | 1.21x | | DOT1L | 1.29x | | AFF1 | 1.36x | **Not one factor reaches the pre-registered 1.5x, so B1 fails - and the failure is mine, not the data's.** The top-decile share is bounded below by 0.1 and above by 1 by construction, and in practice it sits between 0.27 and 0.85 in every track here. A statistic with that little room cannot register the kind of between-sample variation the source paper reports, which it measures with a per-interval oncoprotein score that has no such ceiling [E2]. **There is no spike-in in this deposit, so no unbounded magnitude statistic was available to me** - the depth-free alternatives are all shares. **Recorded as a void test rather than a negative one.** Reporting 'magnitude transfers between infant samples' from these numbers would be a claim about the metric's ceiling. The project's standing 'ordering transfers, magnitude does not' is neither confirmed nor refuted here.