Experiment 37

“Sanborn's triple matrixes: three grid stages in a row, sampled, then the tableau”

Three transposition stages instead of one or two. The space is too large to enumerate, so 200,000 triples are drawn at random from plain columnar widths 2-30, rotations, and keyed columnar from vocabulary words, each stage forward or inverse. The survival test and English scoring of experiments 6 and 16 are unchanged.

revertedstatus
experiment 16parents
3 mintime, all phases
0candidates
0/24known letters
1

The experiment method

What the method does with the 24 known letters.

The artist said he sometimes scrambled the letters through three grids before using the tablethe table The right-hand half of the sculpture, 26 rows of the alphabet each shifted one place further, starting from the word KRYPTOS.. There are over a million ways to pick three grids from our set, too many to try with the slow English check, so we draw two hundred thousand at random. For each, we follow the 24 known lettersthe known letters Sanborn has confirmed that positions 22 to 34 of the message read EASTNORTHEAST and positions 64 to 74 read BERLINCLOCK. through all three grids, ask whether a short secret word could then produce the carved lettersthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. where they land, and whether the letters of that word look like English. Three grids in a row behave like a random shuffle, so the answer should look like the earlier grid experiments.

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 ACCENTshortest key length each sample allows, 200,000 sampled chains of three transposition stages202324252627282930313233shortest: 20 letters, 1 route-5-4-3best on each shuffled copperEnglishbest on the copper: INDERSH, -3.61English score of the best key runOUT: NO READING; THE KEY RUNS ARE CHANCE-LIKEEASTNORTHEASTBERLINCLOCK
the routes

Three transposition stages in a row, drawn from 211 single stages (keyed column shuffles, rotations, grids), sampled 200,000 times, each followed by the table.

the scale

The best key run on the copper scores above every shuffled copper, but the shuffles were run with far fewer samples than the copper, so the comparison is unfair.

the lesson

Recorded as chance, and a rule added to the loop: every diagnostic needs at least ten shuffled controls run at full size, with the same number of trials as the copper.

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. Pick three grids: rows of 7 read down, then a quarter turn of a 12-wide grid, then columns in the order of the letters of KRYPTOS.
  2. Follow position 22, the E of EAST, through the first grid to its new place, then through the second, then the third.
  3. Do this for all 24 known letters.
  4. The carved letter at each final place, with the known letter, forces one key letter on the table.
  5. Two final places a multiple of p apart share a key letter at period p; a disagreement kills p.
  6. The smallest surviving p is this triple's score; write its forced letters into their slots and read them.
the worksheet: the first steps by hand, on squared paper
SCORING THE BEST KEY RUN, INDERSH, BY ITS RUNS OF FOUR LETTERSfour letterslog10 chanceINDE-3.79in the tableNDER-3.34in the tableDERS-3.57in the tableERSH-3.76in the tablemean -3.61Average the logarithm of how often eachrun of four letters appears in English.English text averages about -3.6.INDERSH averages -3.61. A run this shortscores this well by luck somewhere amongtens of thousands of routes; the shuffledcoppers show how often.
4

Worked example, generated

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

stages: columnar 7, then rotate 12, then keyed columnar KRYPTOS; convention kry/kry/kry Vigenere

plain pos  letter  final place  carved  forced key letter
       22    E         88        U       D
       23    A         62        F       O
       24    S         37        W       L
       64    B         26        Q       F
       65    E         97        R       J
       66    R         30        N       M
5

Hypothesis, for a cryptanalyst

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

Sanborn, CNN 2005 (C1): 'double matrixes and triple matrixes... then went back to this table'. Experiments 6, 8, 16 and 26 covered single and double stages exhaustively for their route sets. Triples of 110 stage choices number about 1.3 million, beyond the budget with the English scoring, so this is a sample: 200,000 triples drawn uniformly, with a fixed seed, each stage one of plain columnar at widths 2-30 forward or inverse, a quarter-turn rotation at widths 2-30, or a keyed columnar from a vocabulary word forward or inverse. For each triple under the 8 distinct plain-cipher conventions the forced tableau key letters give the live periods by the divisor rule; the smallest live period is recorded, and the forced key at live periods up to 40 is scored as English on runs of five or more slots against ten shuffled controls. A triple with a live period under 20 or an English key above the controls would be a lead. Bounds: a 15 percent sample, not exhaustive; the smallest-period distribution says whether triples behave as random permutations.

6

Prediction

Written before the run; the result decides it.

The smallest live period over the sample is 22 or 23, as for single and double stages. The best forced key scores inside the range of ten shuffled controls. No candidate.

7

Method

  1. Build the stage list.
  2. Draw 200,000 triples with a fixed seed.
  3. For each triple compose the three permutations and send the crib positions through.
  4. For each of the 8 conventions force the key letters, find live periods by the divisor rule, and score the forced key at live periods up to 40.
  5. Keep the twenty best keys and the smallest-period histogram.
  6. Repeat on ten shuffles.
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
34champion before

Diagnostic, no candidates. 200000 triples x 8 conventions; smallest live period 20; best forced key -3.613 vs shuffled [-4.821, -4.937, -5.151, -4.484, -4.663, -4.922, -4.484, -4.857, -4.623, -4.478].

What was searched, as the harness recorded it.

family
three transposition stages (sampled) then periodic tableau
stages
211
triples sampled
200,000
conventions
8
keys scored
123,792
smallest live period
20
min live period histogram
20
1
23
32
24
492
25
211
26
1,001
27
1,412
28
3,483
29
1,643
30
5,053
31
3,371
32
9,537
33
8,251
best
  1. score
    -3.613
    stages
    keyed ABSCISSAE inv, keyed HILL, keyed VIRTUALLYINVISIBLE
    convention
    kry/kry/kry/beau
    period
    35
    runs
    INDERSH
  2. score
    -3.894
    stages
    keyed NORTH inv, rotate 11, rotate 5
    convention
    std/kry/kry/vig
    period
    39
    runs
    TWEREPA
  3. score
    -4.155
    stages
    keyed PALIMPSEST, columnar 18 inv, columnar 26
    convention
    kry/kry/kry/beau
    period
    36
    runs
    ORCESCO
  4. score
    -4.181
    stages
    keyed SHADOW inv, columnar 12, columnar 18 inv
    convention
    kry/kry/kry/vig
    period
    39
    runs
    DARTHTH
  5. score
    -4.251
    stages
    columnar 15, keyed BERLINWALL inv, keyed CARTER inv
    convention
    kry/kry/kry/vig
    period
    40
    runs
    RORIF, EMONTI
  6. score
    -4.281
    stages
    keyed WEST, rotate 2, keyed BURIED
    convention
    std/std/kry/vig
    period
    40
    runs
    ETINASCR
  7. score
    -4.281
    stages
    keyed WEST, rotate 2, keyed BURIED
    convention
    std/std/kry/vig
    period
    38
    runs
    ETINASCR
  8. score
    -4.281
    stages
    keyed WEST, rotate 2, keyed BURIED
    convention
    std/std/kry/vig
    period
    36
    runs
    ETINASCR
  9. score
    -4.281
    stages
    keyed WEST, rotate 2, keyed BURIED
    convention
    std/std/kry/vig
    period
    34
    runs
    ETINASCR
  10. score
    -4.281
    stages
    keyed WEST, rotate 2, keyed BURIED
    convention
    std/std/kry/vig
    period
    32
    runs
    ETINASCR
shuffled control best
-4.821, -4.937, -5.151, -4.484, -4.663, -4.922, -4.484, -4.857, -4.623, -4.478
control note
controls use the first 6,000 triples of the same sample

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