Four Appleseed Cast tabs, arranged for three guitars

Built 2026-09-07 on MacBookPro from the live Songsterr exports pulled the same day by the chat titled Zappa paren fix audit. Every number below comes from a run of the gate, not from an estimate.

What was asked
What the source files looked like
How the arrangement works
Result, per song
The two tier-1 gates
Subdividing Guitar 3, and what is still open
Comparison with independent transcriptions
The files

700rule-breaking beats before

59rule-breaking beats after

641beats fixed

0pitches lost or invented

1. What was asked

Take the single guitar part in each tab, arrange it across two or three guitars, keep every hand inside five frets, and allow a string skip only when the shape is an octave. Compare the result against other transcriptions online where any exist.

2. What the source files looked like

Every one of the four tabs already carried two guitar staves. Not one beat on any staff exceeded five frets, so the five-fret rule was satisfied on arrival. The other rule was broken 700 times: that many beats attacked two or more strings with an unused string sitting between them, and none of those were octave shapes.

song	staves	beats	beats with a string skip
Secret `s6388915`	2	1502	264
Kilgore Trout `s6389251`	2	1141	90
Blind Man's Arrow `s6389274`	2	1402	294
A Tree For Trials `s6389335`	2	342	52

3. How the arrangement works

A skip means the pick crossed a string it never struck. Two guitarists collapsed onto one staff produce exactly that shape, so the repair gives the notes back to a second pair of hands rather than re-fingering the beat.

Each failing beat is tested for the smallest set of notes whose removal leaves the rest playable exactly where the source wrote them. Those notes move to a third guitar staff. Nothing else is touched, so a healthy beat keeps the string and the fret the transcription gave it.

Three guitars was the ceiling, and three was never needed for any single beat: the solver reported zero beats requiring a third hand and zero beats with no legal shape at all. Guitar 3 exists to receive what two hands cannot hold at one instant, and it carries between 42 and 202 beats per song.

4. Result, per song

song	guitars after	breaking a rule before	after	share fixed
Secret	3	264	25	239 of 264
Kilgore Trout	3	90	1	89 of 90
Blind Man's Arrow	3	294	31	263 of 294
A Tree For Trials	3	52	2	50 of 52
total		700	59	641 of 700

Guitar 3 finished clean on all four songs. Zero skips, zero spans past five frets, zero unplayable beats on the added staff in every file.

Per staff, after the arrangement

song	staff	beats	notes	skip beats
Secret	Distortion Guitar	653	846	1
	Overdriven Guitar	849	1158	24
	Guitar 3 (arranged)	225	242	0
Kilgore Trout	Distortion Guitar	565	649	0
	Distortion Guitar	576	691	1
	Guitar 3 (arranged)	88	89	0
Blind Man's Arrow	Distortion Guitar	716	914	5
	Overdriven Guitar	686	932	25
	Guitar 3 (arranged)	218	268	0
A Tree For Trials	Acoustic Guitar (steel)	175	188	0
	Acoustic Guitar (steel)	167	179	2
	Guitar 3 (arranged)	43	50	0

5. The two tier-1 gates

impossible_gate.py --source BEFORE AFTER was run on every file. Both mandatory gates pass on all four.

song	pitch gate	position gate
Secret	PASS	PASS
Kilgore Trout	PASS	PASS
Blind Man's Arrow	PASS	PASS
A Tree For Trials	PASS	PASS

The position gate reads: every note in a beat that was already healthy kept the string and fret the source gave it. That is the check that catches a blanket refret quietly rewriting a performance while every other number looks fine.

Two traps the build hit and cleared

GPIF Notes are shared definitions. One <Note id> is referenced by many beats, so editing it in place rewrote every beat pointing at it. The first run moved 130 of 205 notes on one staff while only 36 edits had been issued. Copy-on-write brought that to 23.

Beats are shared too. A beat reused at two onsets lost its note in both places when one of them shed it, which showed up as 55 notes placed on Guitar 3 and never removed from the origin. Beat-level copy-on-write took the orphan count to 0 on all four files.

6. Subdividing Guitar 3, and what is still open

An earlier build stopped at 109 remaining beats, and 87 of those 109 failed for one reason: Guitar 3 borrows its bar skeleton from an original staff, so it had no beat at that onset to receive anything. The restaffing chain that normally solves this, table_to_midi.py plus the retime tool, is not on this Mac, confirmed by direct search for all six pipeline scripts.

The other route needs no such tool. Every Guitar 3 beat carrying no shed notes is a rest, and cutting a rest into shorter rests changes nothing that sounds. The missing onset is created by splitting the rest that contains it, and the shed notes land there. A sounding beat is never cut. That took the remaining count from 109 to 59.

59 beats still attack across an unused string, out of 4961 beats across the twelve staves.

reason the beat is still open	count
the rest containing the onset will not divide into plain note values	33
no removal leaves both halves playable	19
Guitar 3 already holds a full hand at that instant	7

The first group wants tuplet-aware subdivision, which is the rhythm red zone where a wrong rational beat position corrupts the bar. That one is parked until it can run against the retime chain.

7. Comparison with independent transcriptions

Songsterr holds exactly one transcription of each of these four songs and all four are ours, so Songsterr offers nothing to compare against. Ultimate Guitar carries three independent guitar transcriptions.

source	song	declared tuning	capo	widest span measured	columns past five frets
UG 578874	A Tree For Trials	D A D G B E	none	0	0
UG 208140	Blind Man's Arrow	none declared	none	3	0
UG 977078	Blind Man's Arrow	none declared	6	22	4

Our A Tree For Trials tuning is independently confirmed. Our file declares pitches 38 45 50 55 59 64, which is D A D G B E. The Ultimate Guitar transcription declares the same tuning in words.

No independent transcription supports capo 7 on Blind Man's Arrow. One declares no capo, one declares capo 6, and our file declares capo 0. The rejected capo 7 that the skill keeps as its standing proof of an ergonomically attractive and musically wrong answer stays rejected.

The human transcribers stayed inside five frets across the material that parsed cleanly, and they wrote string skips of their own. That matches the reading this arrangement takes: a skip is ordinary in a tab, and it often marks a second player.

No tab content is reproduced here. The comparison reads tuning, capo, fret spans and skip counts out of the pages, then discards the notation.

8. The files

Four arranged Guitar Pro files. They open in Guitar Pro, TuxGuitar, or any GP7 reader.

Secret — arranged.gp Kilgore Trout — arranged.gp Blind Man's Arrow — arranged.gp A Tree For Trials — arranged.gp score.json arrange_stats.json index.md

Reproduce it

python3 ~/Projects/_outputs/appleseed-allparts/arrange-2026-09-07/tools/arrange.py
python3 ~/Projects/_outputs/appleseed-allparts/arrange-2026-09-07/tools/refine.py
python3 ~/Projects/_outputs/appleseed-allparts/arrange-2026-09-07/tools/subdivide.py
python3 ~/Projects/_outputs/appleseed-allparts/arrange-2026-09-07/tools/score.py
python3 ~/.claude/skills/impossible-guitar-parts/impossible_gate.py \
  --source _live_export_2026-09-07/s6389274-Blind-Mans-Arrow.gp \
  arrange-2026-09-07/arranged/s6389274-Blind-Mans-Arrow-ARRANGED.gp

Sources. Live tab exports ~/Projects/_outputs/appleseed-allparts/_live_export_2026-09-07/ · arranger and scorer ~/Projects/_outputs/appleseed-allparts/arrange-2026-09-07/tools/ · gate ~/.claude/skills/impossible-guitar-parts/impossible_gate.py · comparison pages UG 578874, UG 208140, UG 977078. Built 2026-09-07.