Experiment 100
0092, 0093 and 0097 each left one cell where K4 sat at or just above the top of ten scrambled coppers: path routes at periods 22-23, one-miss periodic keys at period 22, and the free Quagmire IV at period 3 after a move. Ten controls cannot tell a 1-in-11 event from a signal. Here exactly those cells are recounted, with look-alike routes collapsed into distinct check sets, against fifty new coppers, and each gets an empirical p-value.
What the method does with the 24 known letters.
Three times tonight K4the carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. scored a little higher than all ten scrambled copies of itself. With only ten copies, that happens by luck about one time in eleven, so it proves nothing. Here we redo exactly those three counts against fifty fresh scrambled copies and count, for each, how many copies do at least as well as K4.
Words in the margin are defined on the glossary page.
The carved letters go in at the top; the reading comes out at the bottom.
Four counts in earlier runs sat at the edge of what ten scrambled coppers showed. Here each is recounted as distinct sets of checks, so look-alike routes count once, against fifty fresh scrambled coppers.
Three fall back inside the range, with p of 0.24, 0.48 and 0.10. Quagmire IV at period 3 stays above all fifty: 12 against 0-11, p < 0.02, short of the 0.0125 four tests would need.
It is the same K4 data with a sharper chance level, not new data, and it was the most extreme of 12 cells looked at; corrected for that it is near 0.24. Experiment 0102 tests it directly.
This experiment produced no candidate: the search ruled everything out before a full text was built. Its product is the bounded negative in section 7.
Redo the method on a few letters with pencil and paper.
The same steps run by the code on the real letters.
the four counts replicated, with what ten coppers showed in the original runs
0092 paths, after the move, period 22: K4 13 distinct check sets, ten coppers 8-15
0092 paths, after the move, period 23: K4 88, ten coppers 74-98
0093 one miss, after the move, period 22: K4 980 settings, ten coppers 526-859
0097 Quagmire IV, after the move, period 3: K4 12 distinct equation sets, ten coppers 1-11
here: the same counts as distinct sets, against fifty new coppers
Family, key space, the constraint under test, and the clues it reads.
A replication, not a new search. Cells, fixed before the run from the records: (1) 0092, write along one of 30 grid paths and read along another, periodic key after the move, periods 22 and 23, distinct (check pairs, class) sets; (2) 0093, type-1 library, periodic key after the move with at most one known letter set aside, period 22, distinct (slot partition, class) sets; (3) 0097, type-1 library plus no move, Quagmire IV with unknown alphabets, key after the move, period 3, distinct canonical equation sets. Statistic: K4's count and the fraction of 50 shuffled coppers (seed 100) whose count is at least K4's. With three cells and four counts the multiple-comparison threshold is about p < 0.0125 for any one of them.
Written before the run; the result decides it.
Every cell has an empirical p above 0.05 against the fifty fresh coppers: the excesses were luck. Zero candidates.
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
0092 paths, after the move, period 22: K4 13, coppers 4-20 (mean 11.02), p 0.24; 0092 paths, after the move, period 23: K4 88, coppers 67-116 (mean 88.66), p 0.48; 0093 one miss, after the move, period 22: K4 183, coppers 134-195 (mean 160.94), p 0.1; 0097 Quagmire IV, after the move, period 3: K4 12, coppers 0-11 (mean 4.8), p 0.0.
What was searched, as the harness recorded it.
How this experiment sits among the others is on the lineage page; its family's open hypotheses are on the families page.