Experiment 138

“F1.5 design 1, checked: is the climb's power at key length 23 about English order? Each planted text against scrambled copies of itself”

F1.5, validity check of design 1 before its pass at length 23 is used. 0137 counted a planted text as found when its best climbed score beat every scrambled copy of K4 (10 of 12 at 23). But a planted ciphertext and K4 differ in more than word order: their letter counts differ (planted ciphertexts average an index of coincidence of 0.039, K4 has 0.036, at the 17th percentile of plants), and shuffles of K4 fit length 23 rarely while shuffles of a plant fit 43% of the time. Here each of 12 new planted texts is compared only with 6 shuffles of its own letters that fit length 23 exactly, so the only difference left is the order of the letters. Same climb, same effort as 0137.

revertedstatus
experiment 137parents
31 mintime, all phases
0candidates
0/24known letters
unclocked
1

The experiment method

What the method does with the 24 known letters.

The last experiment's search seemed able to notice hidden Englisha planted message A real English passage, with the 24 known letters written in at their places, enciphered by the method under test with a random key.: planted messages climbedtrial and error Letting a program change a key one small step at a time, keeping steps that improve a score and sometimes accepting worse ones to escape a rut. higher than scrambled copies of the carved textthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page.. But the planted messages and the carved text also differ in how often each letter appears, and that alone might help the search. Here each planted message is compared only with scrambled copies of itself, which have exactly the same letters in a different order. If the search still picks out the planted message, it is reading English word order, and its verdict on the carved text can be trusted at that strength.

Words in the margin are defined on the glossary page.

1.1

The method at work

The carved letters go in at the top; the reading comes out at the bottom.

OBKRUOXOGHULBSOLIFBBWEASTNORTHEASTOTWTQSJQSSEKZZWATJKLUDIAWINFBBERLINCLOCKWGDKZXTJCDIGKUHUAUEKCARIN: THE COPPER, ALL 97 LETTERS, THE 24 KNOWN PLACES IN ACCENTdesign 1 (0137's climb), key length 23: each cell one planted English text; does it beat all 6 scrambled copies of its own letters?123456789101112beats all its own copiesdoes not6 of 12; luck alone gives 1.71; bar 9the same planted texts against 0137's scrambled copies of K4 (other letters, not just another order)12345678910111211 of 12letters right in the best reading of each planted text: 31-43 of 97, 24 of them given; none recoveredOUT: NO READING; DESIGN 1'S POWER CAME MOSTLY FROM LETTER COUNTS, NOT ENGLISH ORDEREASTNORTHEASTBERLINCLOCK
the idea

0137's planted texts beat scrambled copies of K4. But a planted ciphertext and K4 also differ in their letters; the fair comparison is a planted text against shuffles of its own letters.

the check

12 new planted texts at length 23, each with 6 shuffles of itself that still fit the known letters; 0137's climb at full effort on all 84.

the bound

6 of 12 beat all their own copies (luck gives 1.7; bar 9), against 11 of 12 versus K4's copies. Design 1 fails; 0137's verdict on K4 is weak.

2

Best candidate against the known letters

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.

3

Walkthrough, by hand

Redo the method on a few letters with pencil and paper.

  1. Take a passage of English, write EASTNORTHEAST and BERLINCLOCK in at their places, and encipher it with a random alphabet and a random 23-letter key, the Beaufort way.
  2. Shuffle its 97 carved letters into a new order; keep the shuffle only if some alphabet and key could still give the 24 known letters at their places. Make six such shuffles.
  3. Run the same search on the planted text and on its six shuffles, and note each best score.
  4. The planted text counts as noticed if it beats all six; by luck alone that happens one time in seven.
  5. Repeat for twelve planted texts.
the worksheet: the first steps by hand, on squared paper
EACH PLANTED TEXT AGAINST SIX SHUFFLES OF ITS OWN LETTERS: BEST SCORESplantits bestbest own copyrank of 7rank of 13 vs K4 copiesletters right1-4.82-4.9111392-5.02-4.8236353-4.81-4.9311414-4.92-4.9311435-4.90-4.7921406-4.91-4.8231347-4.82-4.9611368-4.81-4.9611399-4.81-4.81213710-4.86-4.95113111-4.86-4.74213812-4.92-4.882133Rank 1 of 7 means the planted textbeat all six shuffles of itself. Byluck that happens one time in seven.Against shuffles of K4 almost everyplant ranks first: those shuffles havedifferent letters, which the climbrewards less.
4

Worked example, generated

The same steps run by the code on the real letters.

key length 23, Beaufort form: K4 in rows of 23; each column is one key slot

  O B K R U O X O G H U L B S O L I F B B W F L
  R V Q Q P R N G K S S O T W T Q S J Q S S E K
  Z Z W A T J K L U D I A W I N F B N Y P V T T
  M Z F P K W G D K Z X T J C D I G K U H U A U
  E K C A R

the seven checks every move keeps (place(carved) + place(known) equal within a column):
  column 0: R+S (24) = M+C (70)
  column 1: V+T (25) = Z+L (71)
  column 2: Q+N (26) = F+O (72)
  column 3: Q+O (27) = P+C (73)
  column 4: P+R (28) = K+K (74)
  column 21: F+E (22) = T+I (68)
  column 22: L+A (23) = T+N (69)
5

Hypothesis, for a cryptanalyst

Family, key space, the constraint under test, and the clues it reads.

C27(c), C1(b) as 0137. Cell: one unknown alphabet on both sides, Beaufort form, key length 23 (7 checks). Climb: 0137's compiled annealing in the set where every known-letter check holds, 1,000 restarts x 100,000 moves per text, full-text quadgram mean. Plants: 12 Carter windows (knowledge/sources/carter-1923-tomb-of-tutankhamen-vol1.txt) with the known letters written in, random alphabet and key (0120's generator, seed 138). Own coppers: for each plant, 6 shuffles of its ciphertext that fit length 23 exactly (0119's decider). Null: if the climb reads no English order, a plant beats all 6 of its own coppers with chance 1/7, so about 1.7 of 12. Bar, fixed before the run: design 1's power at 23 stands if at least 9 of 12 plants beat all 6 of their own coppers; otherwise 0137's pass is withdrawn and F1.1 Beaufort 23 goes back to blocked. Also reported: each plant's rank among its own coppers and among 0137's K4 coppers.

6

Prediction

Written before the run; the result decides it.

At least 9 of 12 plants beat all 6 of their own coppers (the order of English letters is what the climb reads; partial correctness, 11-16 of 73 unknown letters right in 0137, needs order). If fewer, the power claimed in 0137 came from letter counts and F1.1 Beaufort 23 returns to blocked. No plant recovered (90+ letters right). Zero candidates.

7

Method

  1. Draw 12 planted texts at length 23 (seed 138).
  2. For each, shuffle its ciphertext until 6 shuffles fit length 23 exactly (0119's decider).
  3. Run 0137's climb on all 84 texts at 0137's effort.
  4. Per plant: found if its best is above all 6 of its own coppers; rank among them; rank among 0137's K4 coppers.
8

Result and bounds

The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.

0candidates returned
0passed every gate
0/24known letters, best
–lowest loss
nonechampion before

plants beating all 6 own coppers: 6/12 (null 1.7); beating all 0137 K4 coppers: 11/12; recovered 0; bar failed

What was searched, as the harness recorded it.

family
power check of 0137's climb: one unknown alphabet, Beaufort form, key length 23; each plant against shuffles of itself
search
0137's compiled climb, 1000 restarts x 100,000 moves per text, T 0.05->0.002
plants
12
own coppers per plant
6
shuffles tried
188
restarts done
min
1,000
max
1,000
c decryption checked
yes
null rate found
0.143
null expected found
1.71
bar
design 1's power at 23 stands if >= 9 of 12 plants beat all 6 of their own coppers
plants found against own
6
passes bar
no
plants recovered
0
plants found against 0137 k4 coppers
11
rows
  1. score
    -4.822
    letters right
    39
    true score
    -4.395
    restarts
    1,000
    own coppers
    -5.153, -5.147, -5.013, -5.005, -4.956, -4.905
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  2. score
    -5.021
    letters right
    35
    true score
    -4.385
    restarts
    1,000
    own coppers
    -5.137, -5.094, -5.074, -5.035, -4.882, -4.821
    rank among own
    3
    found own
    no
    rank among k4 coppers 0137
    6
    recovered
    no
  3. score
    -4.811
    letters right
    41
    true score
    -4.406
    restarts
    1,000
    own coppers
    -5.218, -5.173, -5.12, -5.084, -4.963, -4.93
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  4. score
    -4.922
    letters right
    43
    true score
    -4.17
    restarts
    1,000
    own coppers
    -5.134, -5.09, -5.068, -5.027, -4.953, -4.93
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  5. score
    -4.9
    letters right
    40
    true score
    -4.141
    restarts
    1,000
    own coppers
    -5.064, -5.05, -5.016, -4.94, -4.933, -4.788
    rank among own
    2
    found own
    no
    rank among k4 coppers 0137
    1
    recovered
    no
  6. score
    -4.908
    letters right
    34
    true score
    -4.249
    restarts
    1,000
    own coppers
    -5.043, -4.99, -4.941, -4.934, -4.88, -4.82
    rank among own
    3
    found own
    no
    rank among k4 coppers 0137
    1
    recovered
    no
  7. score
    -4.819
    letters right
    36
    true score
    -4.159
    restarts
    1,000
    own coppers
    -5.096, -5.075, -5.071, -5.067, -4.982, -4.961
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  8. score
    -4.806
    letters right
    39
    true score
    -4.666
    restarts
    1,000
    own coppers
    -5.117, -5.066, -5.025, -5.02, -4.96, -4.959
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  9. score
    -4.813
    letters right
    37
    true score
    -4.211
    restarts
    1,000
    own coppers
    -5.156, -5.065, -4.878, -4.873, -4.834, -4.807
    rank among own
    2
    found own
    no
    rank among k4 coppers 0137
    1
    recovered
    no
  10. score
    -4.855
    letters right
    31
    true score
    -4.441
    restarts
    1,000
    own coppers
    -5.141, -5.115, -5.051, -5.043, -5.003, -4.946
    rank among own
    1
    found own
    yes
    rank among k4 coppers 0137
    1
    recovered
    no
  11. score
    -4.856
    letters right
    38
    true score
    -4.438
    restarts
    1,000
    own coppers
    -5.277, -5.226, -5.202, -5.108, -4.978, -4.742
    rank among own
    2
    found own
    no
    rank among k4 coppers 0137
    1
    recovered
    no
  12. score
    -4.918
    letters right
    33
    true score
    -4.342
    restarts
    1,000
    own coppers
    -5.142, -5.135, -5.12, -4.997, -4.995, -4.883
    rank among own
    2
    found own
    no
    rank among k4 coppers 0137
    1
    recovered
    no
controls
per plant, 6 shuffles of its own ciphertext that fit length 23 exactly (0119's decider, seed 138)

How this experiment sits among the others is on the lineage page; its family's open hypotheses are on the families page.

Rendered from the record on 2026-10-18 12:26 UTC; updated after every experiment.
view on GitHub