1 hour ago
So I used Claude to run the test and this is a summary:
-test was restricted to the first paragraph from f58v
- used Transliteration: Zandbergen–Landini (ZLZI)
-Full-width lines are those at least 80% of the median line length (about 50 letters), so lines 3, 21 and 22 were excluded.
-Gallows are t, k, p, f, with benched forms (cth, ckh, cph, cfh) counted once at the middle letter. X is the letter position divided by line length
-A streak is 3 or more consecutive lines whose X values fit a straight line within ±0.03 of the line length, with a slope of at most 0.1 per line.
-There were two nulls, 2,000 shuffles each. A pools all words, shuffles them and refills lines to the original letter counts. B shuffles words within each line.
Results:
• The paragraph has 112 gallows points and 24 streaks, with 86 points in streaks. The longest streak runs 6 lines, for example line 15 to line 20 with X drifting from 0.19 to 0.45.
• Null A averages 28 streaks and 98 points in streaks, with longest 5.9. The paragraph falls below the average (P of null ≥ real is about 0.8).
• Null B averages 24 streaks, so the paragraph matches it (P about 0.5).
Notes:
- I understand the concept of the test thanks to Jorge_Stolfi for the explication and the method.
-the refill matches line lengths only to within about half a word, so null A lines aren’t exactly the original widths.
-the test parameters: all can be tweaked for different results but ultimately it is a selection
• Tolerance (±0.03 of the line length): wider counts more streaks as real.
• Maximum slope (0.10 per line): how steeply a streak can lean.
• Minimum length (3 lines).
• Full-width cutoff (80% of the median length), which decided which lines were kept.
• What counts as a gallows, and whether X is measured in letters or words
-test was restricted to the first paragraph from f58v
- used Transliteration: Zandbergen–Landini (ZLZI)
-Full-width lines are those at least 80% of the median line length (about 50 letters), so lines 3, 21 and 22 were excluded.
-Gallows are t, k, p, f, with benched forms (cth, ckh, cph, cfh) counted once at the middle letter. X is the letter position divided by line length
-A streak is 3 or more consecutive lines whose X values fit a straight line within ±0.03 of the line length, with a slope of at most 0.1 per line.
-There were two nulls, 2,000 shuffles each. A pools all words, shuffles them and refills lines to the original letter counts. B shuffles words within each line.
Results:
• The paragraph has 112 gallows points and 24 streaks, with 86 points in streaks. The longest streak runs 6 lines, for example line 15 to line 20 with X drifting from 0.19 to 0.45.
• Null A averages 28 streaks and 98 points in streaks, with longest 5.9. The paragraph falls below the average (P of null ≥ real is about 0.8).
• Null B averages 24 streaks, so the paragraph matches it (P about 0.5).
Notes:
- I understand the concept of the test thanks to Jorge_Stolfi for the explication and the method.
-the refill matches line lengths only to within about half a word, so null A lines aren’t exactly the original widths.
-the test parameters: all can be tweaked for different results but ultimately it is a selection
• Tolerance (±0.03 of the line length): wider counts more streaks as real.
• Maximum slope (0.10 per line): how steeply a streak can lean.
• Minimum length (3 lines).
• Full-width cutoff (80% of the median length), which decided which lines were kept.
• What counts as a gallows, and whether X is measured in letters or words
(8 hours ago)DG97EEB wrote: You are not allowed to view links. Register or Login to view.I couldn't find it so I built it in to Voynich Stats You are not allowed to view links. Register or Login to view.Thank you so much!
You can select by word or individual glyph on a folio

