Zomby Woof, Andy, Drowning Witch, Keep It Greasey, The Central Scrutinizer and Watermelon In Easter Hay, compared bar by bar against professional drum transcriptions. Ghost and staccato marks ignored throughout. 2026-09-05.
The standard used here
One professional transcription outranks the tab. The Songsterr side is either a machine guess with a documented failure mode, or a human tab being checked. A printed chart is not held to a two-of-three vote against a machine.
Chart: DRUM Magazine Groove Analysis, headed @ 0:09, quarter = 113. Tab: revision r41271 by Leonard, 2009, 176 bars, 2461 notes on the drum staff.
Timing would have aligned this to the wrong bars.
0:09 at the tab's own 120 bpm lands on bar 5, and bars 1 to 8 carry 24 closed hi-hats and 4 kicks each with no snare at all. The chart's passage is snare and kick with no hi-hat. Matching on shape instead puts it at bars 9 to 16, where the tab also has no hi-hat. The chart repeats a four bar phrase with snare counts 2, 3, 10, 10; the tab runs 3, 3, 12, 13 then 3, 3, 12, 15. Same figure, twice, in both.
| chart bar | chart kick | chart snare | tab bar | tab kick | tab snare |
|---|---|---|---|---|---|
| sys 2.1 | 4 | 2 | 9 | 4 | 3 |
| sys 2.2 | 5 | 3 | 10 | 4 | 3 |
| sys 2.3 | 4 | 10 | 11 | 4 | 12 |
| sys 2.4 | 5 | 10 | 12 | 4 | 13 |
| sys 3.1 | 4 | 2 | 13 | 4 | 3 |
| sys 3.2 | 5 | 3 | 14 | 4 | 3 |
| sys 3.3 | 4 | 10 | 15 | 4 | 12 |
| sys 3.4 | 3 | 10 | 16 | 3 | 15 |
Kick agrees within one stroke in all eight bars. The snare difference is not reported as a finding: this scan runs at 12 pixels per staff space and the detector returned 31 to 43 heads per bar where a dozen exist, because stems fragment the beams into pieces that pass a notehead test. It needs a higher resolution scan.
Chart: DRUM Magazine Groove Analysis, @ 2:10, quarter = 130. Passage identified three ways: the chart says 9/8 at 130, the tab's bars 67 to 70 are 9/8 at bpm 130, and the album FLAC measures 130.8 bpm by librosa beat tracking over 124 to 146 s.
| eighth | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|
| chart | kick | hat | snare | kick | snare | hat | kick | kick | snare |
| tab | kick | hat | snare | kick | snare | hat | snare | kick | snare |
Eight of nine match. The seventh eighth reads kick in the chart and snare in the tab, across all four bars. The grid came from the nine hi-hat heads themselves at x 97, 138, 187, 229, 276, 316, 363, 407, 451; the four measured kicks at 100, 235, 370, 415 land on eighths 1, 4, 7 and 8.
A second gap in the same bars. The tab writes nine closed hi-hats and zero open, zero pedal there, while using both lanes freely in bars 60, 61, 64, 65 and 66. The chart shows a circled X and an X below the staff.
Chart: the Drumnet full-song drumset chart, which prints a bar number at every system start, 1, 4, 6, 9, 11, 13, 15, 17, 20, 23, so the mapping is read off the page. Drumnet bar 1 is tab bar 1, Drumnet bar 2 is the tab's bars 2 and 3 merged into one 10/16, then Drumnet n is tab n plus 1.
| Drumnet | tab | DN events | tab events | difference |
|---|---|---|---|---|
| 1 to 5 | 1 to 6 | 13, 15, 11, 26, 17 | 15, 12, 9, 25, 17 | +2, -3, -2, -1, 0 |
| 6 | 7 | 12 | 17 | +5 |
| 7 to 9 | 8 to 10 | 15, 16, 19 | 16, 16, 16 | +1, 0, -3 |
| 10 | 11 | 22 | 16 | -6 |
| 11 to 12 | 12 to 13 | 18, 19 | 16, 16 | -2, -3 |
Tab bars 10, 11, 12 and 13 are four identical bars with no hi-hat and no snare.
Each reads exactly floor tom 4, kick 4, low-mid tom 8, sixteen events, four
times over. Drumnet's matching bar carries about 22 events including snare, two
pedal hi-hat strokes written as X below the staff, two grace notes and two accents.
I read that bar at 2.6x rather than trusting the detector.
The tab has zero grace notes across all 115 bars of its drum staff.
Drumnet uses them, twice in that one bar alone, and they are flams. A flam is which hand lands and when.
Bars 14 onward are not compared. Every bar was checked against its own meter: width divided by beats should be roughly constant, and the median is 208 px per beat. Drumnet's bar 14 measures 386, so that segment holds two bars and the barline between them was missed. Everything after it shifts.
Tab: live revision r8852151 by Ben Dibden1, drum staff named "Vinnie Colaiuta", 248 bars, 4948 notes. Three printed sources: the DRUM Magazine Groove Analysis, two Odd Meter Lesson pages, and a complete Modern Drummer transcription.
This tab is rich, and it is the one the moderators saved.
629 ghost notes, 29 grace beats, 17 distinct drum lanes: snare 1423, kick 1311, closed hat 533, open hat 428, bell 282, pedal hat 195, crash 173, low-mid tom 170, floor tom 112, hi-mid tom 109, ride 87, low tom 85, and four more. My 2026-08-29 sweep revision r8768457 on this tab was blocked by a moderator, so the 629 survived. It is the only tab in the set where the damage was stopped before it published.
The lesson pages print the exact notation the sweep stripped.
Both Odd Meter Lesson pages show 19/16 bars whose quiet strokes are drawn as noteheads inside round brackets, with accents above. Published professional practice for a ghost note, printed by DRUM Magazine, and it is what the sweep read as a defect.
The chart is headed Verse 1 @ 0:35, quarter = 134, 19/16. The tab's first 19/16 bar is bar 22 at 37.1 s, and 19/16 covers 161 of the tab's 248 bars. Both lesson pages are also 19/16. Nothing to fix.
CORRECTION, 2026-09-06. I published a disagreement that does not exist.
I reported the chart writing 24/16 at Verse 2 against the tab's 21/16, and called it a gap of twelve sixteenths. That was my error. I read the signature off an 880 px downscaled overview of the page rather than cropping the glyph.
Cropped from the Verse 2 system at native resolution, autocontrasted and enlarged nine times, both signatures read plainly:
| Groove Analysis | tab r8852151 | |
|---|---|---|
| Verse 1 | 19/16 | 19/16 |
| Verse 2 | 21/16 | 21/16 |
Both agree. The tab needs no change on this song.
Independently confirmed: Arthur Barrow, Zappa's bassist and rehearsal director, describes the choruses staying in 4/4 with the verses going to 19/16 and 21/16. Published accounts put the first verse, bridge sections and guitar solo in 19/16 and the second verse in 21/16. 24/16 appears in none of it.
Two mistakes with one cause: reading a time signature off a downscaled image,
then blocking the queue item as "source exhausted" without searching. A single web search
settled what the 180 ppi raster could not. Both are now gates, G33 and G38, in
~/.claude/skills/_shared/no_stop_gate.py.
Sources: donlope.net Joe's Garage notes, DRUM! Magazine, Talking Bass.
Chart: Todd Bishop's groove-of-the-day post, one bar, 531 by 142 pixels. Tab: revision r4658080 by femboy.gardevoir, 92 bars.
This chart removes all lane guesswork.
It prints its own labels down the left edge: hihat, high/mid tom, snare, bass. No inference from staff position is needed, the first time that has been true in this series. Staff lines measured at y 76, 87, 98, 109, 119, spacing 10.8 px.
Chart 16 hi-hat strokes. Tab 8.
Chart heads measured above the top line at x 170, 190, 213, 235, 257, 279, 301, 323, 344, 378, 402, 424, 444, 466, 489, evenly spaced at about 22 px, which is sixteen sixteenth positions with one pair merged by the detector.
The tab's steady groove bars 4, 10, 11, 12 and 13 each read hi-hat 8, kick 6,
snare 1 across eleven slots with durations 8 16 16 16 16 8 8 16 16 8 8.
Across the whole part it holds 741 closed and 17 open hi-hat over 92 bars, an average of
8.2 per bar, so this is the tab's habit rather than one bar's slip.
Vinnie Colaiuta plays the hi-hat in sixteenths and the tab writes it in eighths.
The chart also marks accents on the hi-hat line and an o and a + above it. The tab uses the open lane 17 times in 92 bars and never uses the pedal lane. The kick and snare rows were not separated cleanly at 531 px width, so no claim is made on them.
Chart: the Drumnet five-page chart by BartoRomeo, headed "(1:20) A", quarter 55.47. Tab: live revision r8768414, drum staff 105 bars, 2,091 notes.
Structure matches well.
| chart | tab | |
|---|---|---|
| meter | alternating 4/4 and 5/4 | alternating 4/4 and 5/4 |
| census | 2 bars per system, numbers advancing by 2 | 53 bars 4/4, 52 bars 5/4 |
| tempo | quarter 55.47 | quarter 56.00 |
0.95 percent apart on tempo with identical alternation. Two independent readings of an unusual structure agreeing that closely identifies the same music.
The lane counts are not reportable, and the gate said so.
The chart's staff lines sit at y 270, 275, 280, 285 and 290: 5.0 pixels per staff space. The floor is 15. An eye-read at 6x suggests roughly 16 ride strokes in the 4/4 bar against the tab's 14, and that is deliberately not a finding. One eye-read at 5 px per space is the same shortcut that produced a wrong 24/16 on Keep It Greasey. Settling it needs a scan at about three times this resolution.
The real damage does not depend on the scan.
| revision | author | drum notes | ghosts |
|---|---|---|---|
| r7715683 | Ben Dibden1 | 2091 | 1,317 |
| r8768414, LIVE | me, 2026-08-29 | 2091 | 0 |
| r8908034, queued | me, 2026-09-05 | 2091 | 1,317 |
His 1,317 split 1,316 ride and 1 crash across 103 of 105 bars. My stripping revision published and no moderator caught it. Note count identical at 2,091, so only the marking changed. The song is a ride piece, 1,724 of its 2,091 notes are ride, so those marks are the dynamic shape of the whole part.
The queued restore was checked as a set rather than a count:
1317 of 1317 present, 0 missing, 0 invented. A matching count alone proves
nothing: on Zomby Woof a restore with a non-zero ghost count shared none of the author's
placements.
The two human tabs track their charts. Leonard's Andy tab closely: kick within one stroke across eight consecutive bars, and the same absence of hi-hat in the same place.
The machine tabs fail in a shape. Four byte-identical bars of tom and kick where the chart has snare and hat. Nine closed hi-hats where the chart has open and pedal. Zero grace notes in 115 bars. Kick and snare swapped on an off-beat. None of that is random error; it is a simplification repeated forward.
librosa.beat.beat_track
even where the record cannot settle a single note.