HFGCS · Group 4 long messages · 241 of 60+ characters · 221116 – 261005

Long G4 Messages

Most long G4s are built the same way. Split one at its markers and you get five parts:

Later messages under the same prefix often reuse whole blocks from earlier ones.

Two messages are a match when sliding one along the other lines them up far better than chance at some point. How far it slides is the offset. The score is z = (m − n·p) / √(n·p·(1 − p)), where m is the number of matching characters, n the number of overlapping characters, and p (about 0.032) the chance that two G4 characters agree. A match needs z of 15 or more.

Most matches are direct: the later message took the blocks straight from the earlier one. A few are indirect, matching only because both share blocks with a message in between.

Messages
241

long G4s under 31 prefixes, 221116–261005

Reuse
92 of 210

reused blocks from an earlier long G4 under the same prefix (44%)

Matches
167

all between messages of the same prefix. The best cross-prefix pair scores z 10.7

Offsets
94 of 167

are multiples of 22. None is odd

The layout

Here's an example, the MM-164 aired 261004 at 21:38 UTC, one row per line. The closing markers are in blue. The number on the left is where each row starts, and the one on the right is its length.

261004 21:38z MM-164 · plain, 5 blocks
1
MM
2prefix
3
V4MNSRJISBSI6L5YPHJTAJ
22row A · 18 + X??X
25
IGTJILNINDI2X2AAQEEEE
21row B · 17 + quad
46
SNDNSHHSGSXKPGPGNGUUUU
22block 1 · 18 + quad
68
7NILZIIT3ASPORRLRNLLLL
22block 2 · 18 + quad
90
CPPCSOOKIVACBHHHHHWWWW
22block 3 · 18 + quad
112
SG7F2DD3BIOGSJGWHJGGGG
22block 4 · 18 + quad
134
X75B7222QWEHSDHSQ27777
22block 5 · 18 + quad
156
2SYL5KTFH
9trailer

The markers hold across the set. Among the 203 messages of 100+ characters, the 21st and 24th characters match in 178 (88%, where chance would give about 3%), and 161 (79%) have a quad at 42–45. So a plain message with n blocks comes to:

L  =  PR + A + B + n·blk + T  =  2 + 22 + 21 + 22n + 9  =  54 + 22n

      n = 3, 4, 5   →   120, 142, 164   (the three commonest lengths)

Not every message is plain:

Block layout has the full list with counts.

Block reuse

When a message reuses blocks, it usually takes 3 or 4 whole blocks from a message 2 to 5 long G4s earlier under the same prefix, typically about 20 days before (the median).

The copies aren't exact. In a reused block, the quad and positions 4 and 11 nearly always match the source. The rest of positions 1–5 and 11–13 match 60–91% of the time, and the other ten positions only 8–23%. Kept and changed positions has the chart. Blocks also often land a slot or two away from where they were in the source.

Prefixes change every so often, and nothing has been found carrying over from one prefix to the next: no direct copies, and nothing disguised by a constant letter shift, a 22-character repeating key or a letter-for-letter swap. Details are under Across prefix changes.

Why most match offsets are multiples of 22 94 of 167

A block that moves one slot moves 22 characters, and two slots is 44. That's why most match offsets are multiples of 22.

The ones that aren't are still never odd. Every part that can differ in length between two messages of a prefix has an even length: a block (22), the insert (16), the longer heads (40, 46, 52). Row B (21) and the trailer (9) are odd, but every message has them, so they cancel out. Each off-grid group matches one of those parts:

The 167 matches by how far their offset is from a multiple of 22:

multiple of 2294
22n − 23
22n + 24
22n − 46
22n + 42
22n − 616
22n + 613
22n − 87
22n + 814
22n − 103
22n + 105

None of the 167 matches has an odd offset. Across all 28,920 pairs of long G4s, 13,679 (47%) do.

Built 261008 from SHORTWAVES. A long G4 is a Group 4 message of 60 characters or more, first airing only, without partial captures or disregarded reads. Positions count from 1. Generated by eam_block_pages_.py.