Pre-Registration: Falsification Criteria for the Weekday of the Crucifixion (Wednesday, 25 April AD 31)

This document fixes, in advance and in public, the criteria by which the author would abandon his own position on the weekday of the crucifixion of Jesus, and what he would accept as confirming it. It was written for and sent to Paul L. Schmehl before he ran a formal weighting study of the candidate weekdays, and before any result was seen. The author argues, in "Three Days and Three Nights" (2026), that the crucifixion fell on Wednesday, 25 April AD 31. That position rests on a reading of the duration form in Matthew 12:40 as a measure rather than an idiom; on a passion week containing two sabbaths; and on the fourteenth of Nisan as the day of death. The register is bilateral by design. It states seven falsifiers on the day-axis, each with what would count, what would not, and the specific consequence the author commits to if it is met — including the withdrawal of a chapter and of the claim itself. It rules in advance on the three sources an opponent is most likely to submit, so that adjudication cannot happen after the fact. It lists fourteen searches already carried out that returned against the author's position, so that they cannot be re-run selectively. And it fixes an ordered, exhaustive table of outcomes for the study, quantified over classes of weighting rather than over raw counts, with the commitment attached to each. Two things are stated that cost the author: that his own scoring model, under the premise his correspondent holds, predicts the rival result; and that a favourable outcome on the weekday would not establish the year. The document is deposited to give the registration an independent timestamp, since a pre-registration verified only by its author's word is not a pre-registration. It is offered as an instrument as much as an argument: the procedure is portable to any chronological dispute in which the assumptions carrying the weight are rarely listed. Note on version 3: 260821 Three amendments were made before the study was run; they are described in §1 and marked in §3c, Y5 and Y6. Version 1 remains visible and citable on the same record. Note on version 4: 260906. This version records one research round in full: the arrival of a named, published opponent (A. Reis, 2025) and the primary inspection of the eight sources on which his case against AD 31 rests. Six of those eight I had previously known only through his citation of them, or not at all. Five amendments cost me something. A rule I had leaned on is not in the text it is usually attributed to. A registered crescent value turns out, on an independent model, to sit just below that model's own comfort threshold. An audit of my own source register found six marks claiming more than their entry showed, and one claiming a primary reading for which no record existed. And I made an arithmetical error of my own in this round, caught before dispatch, which is recorded here rather than quietly repaired. The fifth came last: the year-axis support I had described as resting on John 2:20 rests on a preferred reading of a source that contradicts itself, and the precision lost is lost for every candidate year alike. Three returned in my favour, and they are registered with their offsets, as §3c requires. One is neither: a test I ran hoping it would discriminate, which does not, and which I therefore undertake not to use in either direction. The prior versions stand unaltered on the record and remain citable. Nothing has been removed.