Experiment 9

“Do the two crib runs force a readable key, with the copper exact or off by one letter?”

Instead of matching forced key fragments against a word list, score them as English. Add the one-letter-error parameter: a letter inserted or deleted at every position.

revertedstatus
experiment 5parents
2 mintime, all phases
0candidates
0/24known letters
1

The experiment method

What the method does with the 24 known letters.

If the fourth messagethe message The 97 letters of English that Jim Sanborn wrote and enciphered. used a secret word with the sculpture's tablethe table The right-hand half of the sculpture, 26 rows of the alphabet each shifted one place further, starting from the word KRYPTOS., the 13 known letters of EASTNORTHEAST pin down 13 letters of that secret word in a row, and BERLINCLOCK pins down 11 more. A secret word made of English should look like English in those stretches. We read them off under every way the table can be used and ask a simple question: does any stretch look like English rather than noise? We also allow for the sculptor having dropped or added one letter somewhere, as he did in the second message, which would shift everything after that point by one.

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 ACCENTthe key letters the known letters force: plain on the sculpture's alphabet, carved on the plain oneUELRXLONZVDMOFNOEEADIAGKFNOEEADIAGK: -6.15, the best on the exact copper-10-9-8-7-6-5-4-3random lettersEnglishexact copperone letter inserted: FNOEEADIAG -5.79English score of a stretch, 9,408 stretches scoredOUT: NO READING; THE STRETCHES IMPLY NO FULL KEYEASTNORTHEASTBERLINCLOCK
the forced letters

At a key length of 27 or more, the 13 known letters of EASTNORTHEAST and the 11 of BERLINCLOCK each force a run of consecutive key letters. If the key were English words, those runs would read like English.

the scale

Every run was scored under all 24 ways of using the table, with the copper exact or one letter added or dropped anywhere: 9,408 runs. The best, FNOEEADIAG, sits 2.3 standard deviations above random letters and 3.0 below English.

the verdict

That is what the best of thousands of noisy runs looks like. The forced key is not an English phrase on the sculpture's alphabets.

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 the convention: plaintext and key on the sculpture's alphabet, Vigenere.
  2. Position 22 is plaintext E, ciphertext F. In the sculpture's alphabet E is index 10 and F is index 11, so the key letter has index 1: R.
  3. Position 23 is A against L: A is 7, L is 18, key index 11: D.
  4. Carry on through position 34 to get 13 key letters in a row.
  5. Read them aloud. A real key word gives something like a word; noise gives something like RDQXJ.
  6. Score it by counting how common each run of four letters is in English.
  7. Now suppose the sculptor dropped a letter somewhere before position 22: shift the ciphertext one place and repeat.
  8. The best stretch across all of this is the answer; compare it with real English windows of the same length.
the worksheet: the first steps by hand, on squared paper
THE FORCED KEY LETTERS UNDER EASTNORTHEAST, EVERYTHING ON THE SCULPTURE'S ALPHABETplaceknownits no.carvedits no.carved - knownkey22E11F121R23A7L1710D24S6R121U25T4V2218M26N19Q201R27O5Q2015I28R1P32Y29T4R123W30H14N195O31E11G132Y32A7K019N33S6S60K34T4S62Ythe run, read acrossRDUMRIYWOYNKYRDUMRIYWOYNKY scores -7.53 asEnglish.Runs of real English this longscore about -4.25; random lettersabout -8.29.Twenty minutes by hand for one wayof using the table. The run scored9,408 such runs.
4

Worked example, generated

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

forced key stretch under EASTNORTHEAST, convention kry/std/kry Vigenere, exact copper

pos  plain  cipher  key letter
 22    E      F       Q
 23    A      L       T
 24    S      R       E
 25    T      V       L
 26    N      Q       W
 27    O      Q       E
 28    R      P       H
 29    T      R       G
 30    H      N       Z
 31    E      G       U
 32    A      K       P
 33    S      S       F
 34    T      S       H

stretch: QTELWEHGZUPFH   score -7.72 per quadgram (English about -3.6, noise about -6.5)
5

Hypothesis, for a cryptanalyst

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

Under any periodic tableau whose period the cribs allow (27 and above, 0005), EASTNORTHEAST at 22-34 forces 13 consecutive key letters and BERLINCLOCK at 64-74 forces 11, under each of the 24 conventions. If the key is a word or phrase, as Scheidt insists (C5, C6), those stretches must read as English. Schridde 2017 found them gibberish under the plain tableau; 0005 matched them against 62 vocabulary words and K1-K3 windows only. This experiment scores every stretch by quadgram log-probability per letter and compares it with the distribution of random strings and of true English windows of the same length from K1-K3. Then the one-letter-error parameter (family G, C17, C18, the K2 precedent): a letter inserted or deleted at each of 97 positions shifts the ciphertext under every crib past that point by one, giving new stretches. Expected under the null: the best of about 9,400 stretches sits around two standard deviations above the random mean and well below the English mean. Bounds: all 24 conventions, exact copper and every single insertion or deletion, both crib runs, keys at any period of 27 or more.

6

Prediction

Written before the run; the result decides it.

No forced stretch scores as well as a typical English window of its length. The best stretch scores below -4.5 per quadgram, against about -3.6 for English and about -6.5 for noise. Allowing one inserted or deleted letter does not change that. Zero candidates pass.

7

Method

  1. For each convention, force the key letter at each crib position from the plaintext and the ciphertext letter.
  2. Join the forced letters of positions 22-34 into a 13-letter stretch and 64-74 into an 11-letter stretch.
  3. Score each stretch by mean log10 quadgram probability.
  4. Repeat with the ciphertext altered by one deleted letter at each position, then by one inserted blank at each position, which shifts every crib beyond that point.
  5. Calibrate: score all 13- and 11-letter windows of the K1-K3 plaintexts, and 20,000 random strings.
  6. Report the best stretches with their z-scores against both distributions.
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

Diagnostic: no full key is implied, so no candidates. The verdict is the gap between the best stretch and the English distribution.

What was searched, as the harness recorded it.

family
periodic tableau, separate alphabets, period 27+; forced key stretches scored as English
conventions
24
copper variants
196
stretches scored
9,408
calibration mean sd
english 13
-4.251, 0.466
english 11
-4.252, 0.518
random 13
-8.292, 0.98
random 11
-8.295, 1.08
best overall
score
-5.795
stretch
FNOEEADIAG
copper
inserted before 74
convention
kry/std/std/vig
run
64-74
z vs random
2.32
z vs english
-2.98
best exact copper
score
-6.149
stretch
FNOEEADIAGK
copper
exact
convention
kry/std/std/vig
run
64-74
z vs random
1.99
z vs english
-3.66
top ten
  1. score
    -5.795
    stretch
    FNOEEADIAG
    copper
    inserted before 74
    convention
    kry/std/std/vig
    run
    64-74
    z vs random
    2.32
    z vs english
    -2.98
  2. score
    -5.852
    stretch
    FNOEEADIAP
    copper
    inserted before 73
    convention
    kry/std/std/vig
    run
    64-74
    z vs random
    2.26
    z vs english
    -3.09
  3. score
    -5.926
    stretch
    SVICPIFGROWH
    copper
    inserted before 34
    convention
    kry/std/kry/varbeau
    run
    22-34
    z vs random
    2.41
    z vs english
    -3.59
  4. score
    -5.947
    stretch
    SVICPIFGROWF
    copper
    inserted before 33
    convention
    kry/std/kry/varbeau
    run
    22-34
    z vs random
    2.39
    z vs english
    -3.64
  5. score
    -5.99
    stretch
    FQXDESREPO
    copper
    inserted before 70
    convention
    std/std/kry/vig
    run
    64-74
    z vs random
    2.13
    z vs english
    -3.36
  6. score
    -6.012
    stretch
    FNOEEADIAGW
    copper
    deleted 74
    convention
    kry/std/std/vig
    run
    64-74
    z vs random
    2.11
    z vs english
    -3.4
  7. score
    -6.043
    stretch
    JLJODEIGNOSKH
    copper
    deleted 28
    convention
    std/std/std/beau
    run
    22-34
    z vs random
    2.29
    z vs english
    -3.84
  8. score
    -6.061
    stretch
    FNOEEADIABW
    copper
    deleted 73
    convention
    kry/std/std/vig
    run
    64-74
    z vs random
    2.07
    z vs english
    -3.49
  9. score
    -6.077
    stretch
    SYMORCHLKT
    copper
    inserted before 67
    convention
    std/kry/std/vig
    run
    64-74
    z vs random
    2.05
    z vs english
    -3.52
  10. score
    -6.078
    stretch
    PAGIFORWLYA
    copper
    deleted 9
    convention
    std/kry/kry/beau
    run
    64-74
    z vs random
    2.05
    z vs english
    -3.53
exact copper all 24
  1. score
    -6.149
    stretch
    FNOEEADIAGK
    copper
    exact
    convention
    kry/std/std/vig
    run
    64-74
    z vs random
    1.99
    z vs english
    -3.66
  2. score
    -6.151
    stretch
    SVICPIFGROWHF
    copper
    exact
    convention
    kry/std/kry/varbeau
    run
    22-34
    z vs random
    2.18
    z vs english
    -4.08
  3. score
    -6.818
    stretch
    SJRXTUOSORAMN
    copper
    exact
    convention
    std/kry/std/varbeau
    run
    22-34
    z vs random
    1.5
    z vs english
    -5.51
  4. score
    -7.004
    stretch
    QRTPHIUUARAYZ
    copper
    exact
    convention
    std/kry/std/beau
    run
    22-34
    z vs random
    1.31
    z vs english
    -5.91
  5. score
    -7.178
    stretch
    ELYOIECBAQK
    copper
    exact
    convention
    kry/kry/kry/vig
    run
    64-74
    z vs random
    1.03
    z vs english
    -5.65
  6. score
    -7.205
    stretch
    CECHPTSDQDDDE
    copper
    exact
    convention
    std/std/kry/beau
    run
    22-34
    z vs random
    1.11
    z vs english
    -6.34
  7. score
    -7.304
    stretch
    XYHANZEFHYHMK
    copper
    exact
    convention
    kry/kry/std/beau
    run
    22-34
    z vs random
    1.01
    z vs english
    -6.55
  8. score
    -7.383
    stretch
    UGUHMRUKAFK
    copper
    exact
    convention
    std/kry/std/beau
    run
    64-74
    z vs random
    0.84
    z vs english
    -6.05
  9. score
    -7.467
    stretch
    ICXUEILMNSK
    copper
    exact
    convention
    kry/kry/kry/varbeau
    run
    64-74
    z vs random
    0.77
    z vs english
    -6.21
  10. score
    -7.477
    stretch
    JLJODEGKUKKKL
    copper
    exact
    convention
    std/std/std/beau
    run
    22-34
    z vs random
    0.83
    z vs english
    -6.92
  11. score
    -7.532
    stretch
    RDUMRIYWOYNKY
    copper
    exact
    convention
    kry/kry/kry/vig
    run
    22-34
    z vs random
    0.78
    z vs english
    -7.04
  12. score
    -7.558
    stretch
    VNMWWAXSAUQ
    copper
    exact
    convention
    kry/std/std/varbeau
    run
    64-74
    z vs random
    0.68
    z vs english
    -6.38
and 36 more

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