How the Book of Soyga Builds Its Tables and Reads Names from Them (preprint, code and data)

Abstract

Preprint (Author’s Original Manuscript, 14 October 2026; submitted for peer review, not yet peer reviewed), with the code, outputs and working notes, for the article “How the Book of Soyga Builds Its Tables and Reads Names from Them” (2026). The thirty-six letter tables of the Book of Soyga (Aldaraia; British Library, Sloane MS 8; Bodleian Library, MS Bodley 908) were reconstructed by Jim Reeds (1998, 2006) with a step function f that he knew only empirically. Using Jane Kupin’s edition (2014), the scripts test where the book itself states how its tables are made and how names are read out of its squares. Earlier work credited in the manuscript: Louv (2026, doi:10.5281/zenodo.20696278) for the identification of f with the values of Section 18; Ivezaj (2026, doi:10.5281/zenodo.20225605) for the code words and the MOYSES table. Letter values: 112 of the 119 numbers the book attaches to letters or words agree with f(L) + 1 (mod 23); the seven exceptions are listed. Step table (Book 17, Sloane 8, f. 100v): the letter each letter “seeks” is L + f(L) for 22 of 23 letters. Code words from three divine names (Section 28 and a Bodley marginal note): 113–131 of 138 letters, 13–19 of 23 words, against at most 65 letters for 100,000 random pairings. Interior rule: only Reeds’s rule fits his transcription of a manuscript table cell by cell (550 of 560 cells); the failures trace to four wrong letters. Preface spirit names: the spirits of the first table’s preface bear three-letter groups of that table, and those that undo them the same groups rotated (78 of 98 name forms; at most 11 for any other table). The square of Aries (Section 3): a 14 × 14 square from Ihesum, Emanuel, Deus and their reversals; its star names are found within one letter at 17/19, 12/18 and 23/39 (random names at most 6, 5 and 14). Lists 1–2 follow a start-and-step rule; the first 21 names of list 3 are the upper half read along the diagonals from the corner that “sounds deum”; 8 of the last 18 are paths from the main diagonal. Negative results are included: no “19 code” or Metonic constants; the Section 16 trigrams are not taken from the poem; the “alphabet” of Zadzaczadlin is unresolved. All readings rest on Kupin’s transcription; the manuscripts were not inspected. Python 3, standard library only. See README.md and RESULTS.md in the archive.