# Page 23 / Vumarii 2019 positional-alignment audit

## Scope

A bounded test of a position-for-position, simple monoalphabetic Page-23/Vumarii relationship. It does not test arbitrary ciphers, astronomical readings, or thematic relationships.

This is intentionally much narrower than a general cipher search. It checks whether the shared 154-slot count supports a directly aligned, one-symbol-to-one-letter model.

## Inputs

- Page 23: `archive/solve/page23_canonical.json` (SHA-256 `e9047bc912167f7e9e42d589cca300bcf22fcd648a6153fb0a022dd722d0056b`)
- Vumarii transcription: `archive/gaps/tr1-00-files/vumarii-2019-TRANSCRIPTION.txt` (SHA-256 `2c6eb4492a3975c034e9fb4ba75ae676acad0fc979b76e7a0463bd77170ad6b3`)
- Page 23 grid: 11 x 14; aligned slots: 154
- Vumarii transliteration coverage used by the direct test: 134 slots.

## Observed maximum across the eight grid orientations

- Orientation selected by the functional score: `transpose_mirror_tb_14x11`.
- Best glyph-class mapping explains 45 / 154 positions (29.2%); 109 aligned positions conflict with its best per-class mapping.
- Orientation selected independently by direct equality: `transpose_rotate_180_14x11`; exact equality at provisionally transliterated slots: 12 / 134 (9.0%).

## Matched shuffled controls

- Functional mapping: observed 45; controls mean 45.59, range 41-53; 3656 / 5000 controls were at least as large; empirical p >= 0.7313.
- Direct equality: observed 12; controls mean 12.06, range 7-21; 2927 / 5000 controls were at least as large; empirical p >= 0.5855.

## Conclusion

Neither predeclared statistic is unusual under the matched shuffled controls. The shared length therefore supplies no evidence for this simple indexed-substitution family.

A negative result does not prove that Page 23 and Vumarii are unrelated. It does mean that claiming a decode from the common count, without an additional source-selected rule, would not be justified.