Inca Roads: what the drum tab is missing

Songsterr s412178 · live revision r8769026 · Frank Zappa, One Size Fits All, 1975 · built 2026-09-05

165ghost notes stripped and still absent

166dynamic marks added that no author wrote

8bars carry no drums on either staff

0professional drum charts of this song on the drive

1bar where the two drum staves overlap

165/165repair ghosts land in the author’s exact slots

The ask

Compare the live Inca Roads drum tab against every transcription we hold on this Mac and report what is missing or wrong. The comparison ran against the Songsterr revision chain, the archived professional transcriptions at ~/Projects/_outputs/zappa-drum-sources/, and the baseline Guitar Pro export kept for this song. Every number below came from the part JSON that a reader actually receives from CloudFront.

Headline. The single largest problem is my own doing. Revision r8769026, published under Brandon’s account on 2026-08-30, stripped 165 ghost notes from the Chester Thompson staff and added 166 dynamic marks that the previous author never wrote. The repair revision r8907511 is sitting in the moderation queue and has not gone public, so the tab a reader opens today is still the damaged one. That repair was verified at placement level for this page and its ghost restoration is correct.

What this tab is

SongInca Roads
One Size Fits All, 1975

Tabs412178, live revision r8769026
12 tracks, two of them drums

OriginaiGenerated: false
initial revision r279379 by enikey87, 2016

Length322bars, with 42 meter changes

Drum staff 11Chester Thompson
1,529 notes, 966 views

Drum staff 10AI
1,844 notes, 92 views

The revision chain

Revision	Date	Author	Tracks	State	What it says
r8907511	2026-09-05	Brandon chavez	12	on moderation	Restores the 165 ghosts. Not public.
r8769026	2026-08-30	Brandon chavez	12	LIVE	Removed the 165 parenthesised noteheads.
r7750859	2026-07-06	Darr	12	superseded	Reverted a track deletion. Last clean state.
r7723273	2026-07-04	e	10	superseded	Finished the guitar solo, deleted the AI tracks.
r7244427	2026-06-07	Ben Dibden1	13	superseded	Earlier full build.
r279379	2016-09-03	enikey87	8	superseded	Initial revision.

Eleven revisions exist in total. The five omitted rows are earlier Ben Dibden1 and “e” passes that no finding depends on.

Missing: 165 ghost notes

Comparing Darr’s r7750859 against the live r8769026, beat by beat, on the Chester Thompson staff. Beat count and note count are identical at 1,491 and 1,529, so no playing was deleted. What vanished is the ghost flag on 165 of those notes, the bracketed noteheads that mark the quiet strokes between the loud ones.

Lane	MIDI	Ghosts lost	Share
Acoustic snare	38	134	81.2%
Crash cymbal 1	49	14	8.5%
Bass drum	36	7	4.2%
Crash cymbal 2	57	6	3.6%
Closed hi-hat	42	3	1.8%
Outside the kit	92	1	0.6%
Total		165	100%

Why this counts for more than formatting. A ghost on the snare is a real stroke played quietly. Removing the bracket tells a reader to play 134 quiet snare strokes at full volume. The crash ghosts are a separate case worth naming, because a bracketed crash is unusual notation and it belongs to the author who wrote it.

Wrong: 166 dynamic marks that no author wrote

The same sweep added a dynamic to every beat it touched. In r7750859 those 165 beats carry velocity: null. In the live revision they carry 83 pp marks and 83 mf marks. The stated reason was to preserve the quiet reading after the brackets came off. The effect is that a reader now sees written dynamics that Darr never put in the score, on a staff that previously had none.

Pre-sweep r77508590dynamic marks on the drum staff

Live r876902616683 pp and 83 mf

Ghost flags remaining0on both drum staves

The two drum staves divide the song between them

This is the finding that would break any audit that looked at one staff alone. The AI staff carries bars 1 to 155 and is completely empty from bar 156 to the end. The Chester Thompson staff is the mirror image, silent through most of the first half and carrying everything from Interlude 2 onward. Exactly one bar, bar 3, has notes on both.

Staff 11, Chester Thompson · 322 bars

Staff 10, AI · 322 bars

IntroVerse 1Break 1Verse 2Verse 3Interlude 1Guitar SoloPost-SoloInterlude 2Verse 4BreakdownBridgeKeys SoloInterlude 3BoogerOutroEnding

Note density per bar: empty 1 to 4 5 to 10 11 and up

Staff	Bars with notes	Empty bars	Notes
10, AI	1-28, 37-51, 63-137, 141-155	191	1,844
11, Chester Thompson	1-3, 29-36, 52-62, 138-140, 156-289, 296-322	138	1,529

Read this ahead of judging either staff. Neither one is a complete drum part. Anyone measuring staff 11 on its own would report 138 empty bars and call the tab half finished. The correct unit of judgement is the pair.

Missing: six bars with no drums anywhere

Combining both staves leaves eight bars with zero drum notes. Six of them run consecutively and land on a named section.

Bars	Section marker	Staff 10	Staff 11	Verdict
290-295	Interlude 3	empty	empty	six-bar hole
31	Break 1	empty	empty	single bar
33	Break 1	empty	empty	single bar

The two single bars inside Break 1 are plausible as written rests, because a break is where a band stops. Interlude 3 is the one to check against the recording, since six bars of silence going into the Booger section is a large claim for a tab to make.

Wrong: the AI staff was touched after all

The live revision description states that the second drum staff “had none and was not touched.” The beat counts disagree. Staff 10 went from 1,475 beats to 1,501, a gain of 26 across 17 bars. Every added beat is a rest and the note count holds at 1,844, so no playing changed. The claim in the description is still wrong and should be corrected in any future note.

bars affected: 168, 173, 175, 177, 178, 179, 180, 186, 187, 188, 296, 298, 300, 301, 302, 308, 309

Most likely cause. A Guitar Pro round trip re-split rests inside odd-meter bars. All 17 bars sit in the 11/16, 8/16 and 5/16 stretches, which supports that reading and does not prove it.

The note outside the kit

Bar 1 of the Chester Thompson staff carries a note at MIDI 92 on string position -0.5, which places it on a ledger above the staff. General MIDI percussion runs from 35 to 81, so 92 sits outside the drum map entirely and will not sound as any kit piece.

This marking belongs to Darr and should stay. The same note is present in Darr’s r7750859 carrying ghost: true. The marking belongs to the author and predates the sweep. The repair restores its ghost flag and leaves the pitch alone, which is the correct call under the rule that restoring means reproducing what the author wrote.

What we actually hold on the drive

The honest answer to “compare it against all our transcriptions” is that no professional drum transcription of Inca Roads exists in the archive. Every folder was checked.

Source folder	Transcriber	Holds	Covers Inca Roads
02-drummagazine	Ryan Brown	7 notation PNGs	No
01-cruiseshipdrummer	Todd Bishop, Daniel Bédard	Flower Punk, Apostrophe, Yellow Snow, others	No
03-petesweeney	Pete Sweeney	Vinnie transcriptions PDF, 9 pages	No. Dong Work For Yuda and Keep It Greasy only.
06-drumnet	Drumnet.ru	Not recovered on this Mac	No
08-analysis/zappa-analysis	Kasper Sloots	inca-roads-1973-a.jpg, -b.jpg	Melodic only
10-academic	Clement, Wall	Dissertation and thesis	No drum notation

The Sloots charts do not vote on drums. Both images were opened and read. They are handwritten full scores carrying flute, trombone, keyboards, whistling and bass guitar staves, with no drum staff on either page. They also transcribe the 1973 Roxy-era performance in 12/16, and this tab transcribes the 1975 album version that opens in 4/4. They are a separate arrangement of a separate take. Under the rule that a source votes only on the passage it transcribes, they vote on nothing here.

That leaves the Songsterr revision chain as the only valid comparator, which is why every finding above is a revision-to-revision measurement rather than a chart reading.

Is the queued repair correct

The repair revision r8907511 was checked at placement level rather than by counting, because a total cannot see where a mark landed. On Zomby Woof a queued repair of mine held 8 ghosts against the author’s 32 and not one of the 8 sat on a bar and drum the author had marked. The same check was run here, building the set of bar, voice, beat, string and lane tuples on both sides and taking the difference.

Check	Darr r7750859	Repair r8907511	Difference	Verdict
Bar count	322	322	0	match
Ghosts on staff 11	165	165	0	match
Ghosts in the author’s exact slots	165	165	0 missing, 0 extra	match
Dynamic marks on staff 11	1	1	0	the 166 are gone
Staccato flags	1	1	0	match
Beats on staff 10	1,475	1,501	+26	still off

The ghost restoration itself is correct. All 165 land on the exact bar, voice, beat, string and lane that Darr marked. Zero are missing and zero are invented. The 166 fabricated dynamics are reverted. The Zomby Woof check failed to produce this outcome, and the result here was measured rather than taken from the revision description.

One claim in the repair is overstated. Its description says the revision “returns the tab to revision r7750859 exactly.” Staff 10 still carries 1,501 beats against the author’s 1,475, so the 26 rest beats were never rolled back. Note count holds at 1,844 on both, so no playing is affected. The wording needs correcting while this revision still sits in moderation.

The baseline on disk cannot be trusted

Two files sit in the Inca Roads work folder. They are byte-identical, sha256 ddb5235078c6cae4 for both, despite one being named for the original and one for revision r7750859.

File	Bytes	AntiAccent	On the drum staves
ORIGINAL-Frank Zappa-Inca Roads-07-06-2026.gp	111,487	14	zero
RESTORE-r7750859.gp	111,487	14	zero

All 14 ghost markings in those files sit on pitched notes, MIDI 69 and 62 at octave 5, which puts them on the guitar and marimba staves. The drum staves carry none. A file that genuinely held r7750859 would carry 165 on the Chester Thompson staff. Both files are post-sweep exports wearing a pre-sweep name, and the modification date of 2026-08-29 19:27 matches the day the sweep ran.

Operational consequence. Anyone who reaches for RESTORE-r7750859.gp to rebuild this restore would publish a stripped tab a second time. The working repair was built from the CloudFront part JSON instead, which is the method the skill already prescribes. These two files should be renamed to say what they are.

What is not proven

Whether the six silent bars at Interlude 3 are correct has not been tested against the recording. That needs the audio and a separated drum stem, and this pass did not run one.
Whether the surviving hi-hat and cymbal writing on staff 11 is complete is untested. Staff 11 carries 93 hi-hat events against staff 10’s 361 over the same 322 bars, and the two staves cover opposite halves of the song, so that gap is not yet evidence of anything.
The cause of the 26 added rest beats is inferred from where they fall, not measured.
No claim is made about the accuracy of the notated rhythms themselves. The comparison here is between revisions of the same tab, and both sides of it share whatever the original got wrong.

Sources

Live tab, songsterr.com s412178, revision r8769026, part JSON read from CloudFront on 2026-09-05.
Pre-sweep revision r7750859 by Darr, same route, image hash v0-3-2--J1WPCKhadNeBx-S.
Revision list, api/meta/412178/revisions, 11 revisions.
Archived transcriptions, ~/Projects/_outputs/zappa-drum-sources/, manifest ARCHIVE-STATE.md.
Kasper Sloots, zappa-analysis.com, Inca Roads 1973 sections 1 and 2, transcription deposited at the I-depot, The Hague, June 2014.
Damage log, ~/.claude/skills/songsterr-drum-repair/reference/damage-log.md.